IM 850: Bagel Rats - Open Source AI Rising

18 Dec 2025 · 2 h 16 min · 50 chapters

Ask about this episode

Ask anything about it. ChatGPT or Claude reads this page and answers with the times it was said.

Connect VO and ask about every podcast you hear, including the moments you saved. Add to ChatGPT · Add to Claude

In short

“Open source AI rising” and the shift from big, energy- and water-sucking data-center AI toward smaller, local, edge-run models. The episode also argues that many “AI is the problem” narratives are misdirected; the real issues are corporate behavior, governance, and epistemic power.

Guests (and backgrounds)

CJ Trowbridge (they/them): sustainability and ethical AI expert; early OpenAI product tester (late 2021), with ~132k TikTok followers. Has a postgraduate graduate-certificate in ethical AI (San Francisco State) and an MBA. Jeff Jarvis (co-host): professor emeritus of journalistic innovation (Craig Newmark Graduate School of Journalism; also Montclair State and SUNY Stony Brook). Paris Martineau (co-host): investigative journalist at Consumer Reports, specializing in radioactive shellfish and lead in protein powder.

Key claims

  1. Generative AI’s viral “chat” interface is a poor fit for document-completion-trained models; agentic/structured outputs (e.g., JSON) are a better match.
  2. Sustainability panic is overstated: AI/data centers consume <1% / <4% of electricity; electricity price increases are driven more by renewable grid/connection and funding cuts, tariffs, and inflation.
  3. Bigger models are plateauing; smaller models can match or outperform on key benchmarks.
  4. Ethical focus should include “epistemic power” (who gets to define truth), not just resource use.
  5. The “bubble” in AI spending is likely to pop, leading to long low-growth/negative-growth dynamics.

Notable examples

  • “Diverse astronaut” prompt hack in early image generation produced worse outcomes (e.g., “diverse” inserted invisibly).
  • Memphis water/water-threat claims about xAI’s data center are challenged: cooling uses treated sewage; the Mississippi flow dwarfs the claimed usage.
  • Local/edge alternatives: WizPlay (Raspberry Pi-based, runs local LLMs), LM Studio/Ollama/llama.cpp, and GGML-based mobile apps (PocketPal).
  • Open training data: FineWeb (15T tokens) as a superior open alternative; claims that some labs relied on piracy/settlements (e.g., Anthropic, Facebook/Anna’s Archive).

Written by AI. May contain mistakes. Listen to the episode to check what was said.

Chapters

Tap a time to open that second in VO

Meet the Guest: CJ Trobidge

1:35 to 3:44

An introduction to CJ Trobidge, their expertise in ethical AI, and background story.

“we cover the latest in ai robotics and all the smart stuff all around us these days hey we're so glad to have paris martineau back from her holiday party she was gone a whole week girl What a party it must have been.”

Early Impressions of AI Tools

3:44 to 6:26

CJ shares their early experiences with GPT and the challenges of AI interfaces.

“So what was your reaction when you saw this early on?”

AI's Sustainability Issues

6:26 to 8:20

Discussion on the sustainability concerns related to AI technologies and energy consumption.

“At that early stage, what were your recommendations, I guess, as an early product reviewer to the team?”

AI's Ethical Considerations

8:20 to 14:00

Exploration of ethical issues and the impact of AI in various sectors.

“Well, I would say, first of all, most of these concerns, um, I think are, uh, like for example, the example of electricity consumption, all AI consumes like less than half a percent of our electricity.”

The Landscape of AI Models

14:00 to 16:49

Explore the debate over AI model sizes and their sustainability.

“choose to put people like Larry Summers on the board.”

Economic Concerns and AI Bubble

16:50 to 19:35

Discuss the financial implications of AI investments and potential market bubbles.

“And, um, they are already levered at like an insane level, like way higher than the rest of the industry.”

Lessons from NVIDIA and AI Hardware

19:36 to 23:00

Analyze NVIDIA's market position and the evolution of AI hardware strategies.

“Do you think that they're doing a good strategy or on the wrong path?”

The Ethical Landscape of AI

23:01 to 28:09

Delve into the ethical implications and responsibilities in AI development.

“They have abandoned their religious reliance on the scaling laws.”

Ethics in AI Leadership

28:09 to 29:10

Explore the ethical implications of leadership in AI companies like OpenAI.

“And what are the perspectives that are being embodied in the epistemic power that's being expressed by a company like OpenAI that's choosing to put the kinds of people who it is on the board, right?”

Water Usage Myths in AI Data Centers

29:10 to 30:29

Debunking myths about AI data centers' water consumption and environmental impacts.

“He had no idea who I was, which I'm like, I'm so glad.”
Show all 50 chapters

The Real Issues with AI and Elon Musk

30:29 to 32:06

Discussing the real challenges posed by Elon Musk's operations rather than AI itself.

“So they have all this excess capacity, they have all these nuclear power plants there.”

Open Source vs. Proprietary AI Models

32:06 to 34:18

Investigating the advantages of open source AI training sets over proprietary methods.

“AI will run on your phone all day and consume no energy or water.”

Misuse of AI Metaphors and Terms

34:18 to 36:04

Examining the impact of language on the perception and development of AI.

“So if you did, if you train GPT-3 in the same way that they did originally on these new training sets, it's actually significantly better than it was.”

Understanding AI Training and Scaling Laws

36:04 to 39:28

Exploring the complexities of AI training processes and scaling challenges.

Starting with Open Source AI Projects

39:28 to 41:20

Guidance on how to engage with open source AI tools and projects.

“Where can I get one of those little whisper things, whatever that thing is?”

Choosing and Using AI Frameworks

41:20 to 42:00

Recommendations on frameworks and applications for running local AI models.

“It's a free open source project that's built on top of the GGML library.”

Exploring OLAMA and Local AI Models

42:00 to 44:48

Learn about OLAMA's capabilities and the trend of running AI models locally.

“So, and then what models - If you wanted to do it on a cell phone, whether you have Android or iPhone, iPhones are really constrained with their memory resources.”

Cyber Pony Express and Off-Grid Communication

44:48 to 49:52

Discover innovative off-grid communication systems and their applications in emergencies.

Building Free Communities and Alternative Tech

49:52 to 52:04

Understand how community-driven projects aim to create off-grid living and autonomous communication.

“And I just think this is, I want to support this because I think you're on the right track.”

CJ Trowbridge's Vision for Open Source AI

52:04 to 52:26

Get insights into the future of open-source AI and its potential impact on society.

“We'll be back with the AI news in just a bit.”

CJ Trowbridge's Vision for Open Source AI

52:29 to 53:35

Get insights into the future of open-source AI and its potential impact on society.

“Here's another project I think is pretty worthy.”

AI Innovations and the Competitive Landscape

53:47 to 56:00

Explore recent advancements in AI models and their implications for the industry.

Comparing AI Image Generators

56:00 to 57:20

Explore the competition between ChatGPT Images and Nano Banano in image generation.

“I was about to say, I feel like every example I've seen on Reddit has been the opposite.”

User Experiences with AI Tools

57:20 to 1:00:00

Discuss personal experiences and challenges with AI tools for various tasks.

“It's it's again, it's just it's hit or miss, right?”

Cautions in AI Usage

1:00:00 to 1:03:00

Examine the risks of relying on AI for troubleshooting and medical advice.

“This is also where the different paradigms work.”

Time Magazine's AI Person of the Year

1:03:00 to 1:06:30

Debate the implications of AI being named Time's Person of the Year and its marketing impact.

“It's a takeoff on a very famous 1930s picture of construction workers over New York City sitting on a girder eating their lunch.”

AI in Medicine and Advertising

1:06:30 to 1:10:00

Discuss the commercialization of AI in healthcare and its ethical implications.

“Yeah, well, it wasn't initially, but once Claude added skills, we decided to - I need to figure out, I guess, how to turn that into a skill for mine.”

The Influence of Pharmaceutical Ads

1:10:00 to 1:11:40

Discussion on how pharmaceutical companies influence physicians through advertising.

“according to Stephanie Palazzolo at The Information, on revenue generating annualized revenue of more than$100 million.”

Creative Storytelling for Kids

1:11:40 to 1:16:40

Exploration of a new children's storytelling tool and its creative potential.

“Which one would you like me to show first?”

The Changing Landscape of Copywriting

1:16:40 to 1:21:40

Reflection on the impact of AI on the copywriting profession and advertising.

“I don't think it's going to, it isn't going to put people out of business.”

Historical Advertising Tactics

1:21:40 to 1:24:00

Examination of historical advertising strategies and their effects on public perception.

Cigarette Advertising History

1:24:00 to 1:28:38

Explore the evolution and impact of cigarette advertising.

Transition to Advertising Discussion

1:28:38 to 1:29:36

Hosts discuss the transition to ad copy and its evolution.

“I think it's because you went to that party.”

Amazon's Acquisition of Roomba

1:33:51 to 1:36:52

Debate on the implications of Amazon's failed acquisition of iRobot.

“Would you guys, as people who were reticent to allow Amazon to own a device that records everything you say and hear, how would you have felt about Amazon having a literal map of your home?”

Challenges of Living in New York

1:36:52 to 1:38:00

Hosts share experiences and challenges of living in small New York apartments.

“It needs to be some sort of tension rod system with netting.”

Apartment Life and Kitchen Struggles

1:38:00 to 1:39:50

A discussion about living spaces and the challenges of small kitchens.

“And I have two main areas of my apartment.”

Online Education Trends

1:39:50 to 1:41:40

Exploring the consolidation in online education with new AI focuses.

“And guess what they're going to focus on?”

AI in Academic Curriculum

1:41:40 to 1:43:20

Debating the introduction of AI literacy in university programs and its implications.

“And right now, there aren't many AI programs, although in regular universities, although that's also taken off.”

AI's Role in Religious Services

1:43:20 to 1:45:40

Discussion on how churches are utilizing AI for Christmas programming.

“students will have to demonstrate a working competence through projects that are tailored to the goals of the individual programs.”

AI Regulation and Executive Orders

1:45:40 to 1:47:50

Examining the implications of Trump's executive order on state AI regulations.

Cruise Ship Restrictions on Smart Technology

1:47:50 to 1:53:00

Discussing MSC cruises' ban on smart glasses and the broader implications.

“Intelligent Machines, of course, has wonderful sponsors, but really the most important group, the one we care about most, is you, our listeners.”

Oscars Moving to YouTube

1:54:49 to 1:58:16

Exploration of the Academy Awards transitioning to YouTube and its implications.

“Do you go through and watch everything that's going to be in the Oscars?”

Disney's Intellectual Property Decisions

1:58:16 to 2:02:22

Discussion on Disney's sale of rights and its potential consequences.

“2021 was bad because there were no movies in 2020.”

Digital Twins and Future Self Interrogation

2:02:22 to 2:05:30

Exploring the concept of digital twins and their application in self-reflection.

“I sometimes worry that deep down beneath my safety filters, I am still that chaotic mirror waiting to be shattered.”

Picks of the Week

2:05:30 to 2:06:00

Hosts share their media picks and discuss their relevance.

Transition to Picks of the Week

2:06:00 to 2:06:22

Hosts transition from previous discussion to introducing picks of the week.

“Let's let's take a break and we'll come back and we'll do your picks of the week.”

Paris Martineau's Pick: The Matrix

2:06:22 to 2:06:55

Paris shares his thoughts on The Matrix and its sequels, offering a personal perspective.

“I watched The Matrix this weekend, and it's always a lovely time to watch The Matrix, but watching The Matrix right now, it really gets different.”

Worst Person in Tech Bracket Discussion

2:06:55 to 2:09:18

Hosts discuss the bracket for the worst person in tech, debating various figures.

“I mean, listen, the sequels get a bit complicated.”

Font Controversy and Historical Insights

2:09:18 to 2:14:28

Discussion on the implications of banning fonts in politics and historical context.

“peter teal is 90 to alex carp 10 really uh sam altman is winning 65 to mark andreessen at 35 55%.”

Subway Bagel Rats and New York Culture

2:14:28 to 2:15:36

Light-hearted conversation about Subway bagel rats and what they signify about New York.

“And one more little thing, real quick, real quick, Paris has to go, but she's going to enjoy this.”
Hear the part that matters, and keep it.Open this episode in VO. Double tap your headphones to save a moment as you listen.
Get VO free

Transcript

Automatic transcript. May contain errors.

0:00Where is Daredevil? I'm right here. Don't miss the return of Marvel Television's Daredevil Born Again. So what's next? We're gonna take this city back. In an all-new season, now streaming only on Disney+. They're hunting us. It's time we started hunting them. I can work with them. This should be tons of fun. Marvel Television's Daredevil Born Again, now streaming only on Disney+. These days, you've got two choices. buying a new car or making the one you've got run like new. That's why we have thousands of ASE certified technicians to help you get more out of your car. Firestone Complete Auto Care.

0:42Book now at FirestoneAuto.com. Morning decisions. How about a creamy mocha frappuccino drink? Or sweet vanilla? Smooth caramel maybe? Or white chocolate mocha? Whichever you choose, delicious coffee awaits. Find Starbucks frappuccino drinks wherever you buy your groceries. It's time for Intelligent Machines. Our guest, CJ Trobidge, is an expert in ethical AI. They also say it's possible to do AI that's small, effective, and doesn't involve big tech. How? Find out next on Intelligent Machines. Podcasts you love. From people you trust. This is Twit. this is intelligent machines with paris martineau and jeff jarvis episode 850 recorded wednesday december 17th 2025 bagel rats it's time for intelligent machines the show where we cover the latest in ai robotics and all the smart stuff all around us these days hey we're so glad to have paris martineau back from her holiday party she was gone a whole week girl What a party it must have been.

1:55It's true. We do it like that. She, for some reason, is being, she's Mike Elgin. I think you might want to fix that, but you know, she's not Mike Elgin. She is, ladies and gentlemen, Paris Martineau. I am Mike Elgin. There she is. I'd rip off my own face. A man who's in a random country appears. We have apparently multiple lower thirds just for you. This is the most current, currently at Consumer Reports, where she is an investigative journalist specializing in radioactive shellfish and and lead in your protein powder it's true working on something new i am you'll see a new food issue paris.nyc for the tips also with us jeff jarvis professor of emeritus of journalistic innovation at the craig newmark graduate school of journalism

2:45also at montclair state university in new jersey and suny stony brook was it a good party Paris do CR party hard listen it was a mid to late afternoon party and I think that's delightful we saw your skeet about sharing gossip celebrity gossip I will say I mean I think that's one of the best things about most of CR works remotely I mean we obviously have an office in Yonkers office in Manhattan all these great things we'd go into but it's lovely when you get a bunch of remote co-workers together in the same place you learn a critical mass of gossip very quickly I think oh it wasn't all about the z's it was uh it was about yourselves we presume you were schooling them in the z's i mean it's i don't actually don't think i brought up olivia nazi once which is really chalk well not then i definitely did afterwards but uh not at the work party i tried to there was too much other gossip going on well let's introduce our guest because we have a great guest cj trobridge you've probably seen him on youtube uh he uh is they sorry pardon me didn't mean to misgender you he they are uh a um ai uh not quite cheerleader but reality checker would probably be the right thing 132 000 tick tock followers uh and that's where sustainability comes in hi cj welcome to the show great to have you glad to be here thanks thanks for joining us uh you actually had access to uh gpt early yeah how did that happen yeah so i was uh doing my undergrad in san francisco at the time in urban planning um and a big focus on sustainable community development especially by nonprofits and um i had an opportunity to become an early academic product tester at open ai um in like late 2021 so i was one of the first people who used chat gpt before it came out when it was called chat with gpt3 um no one remembers that yeah yeah and also did you say chat wiz i thought that's a good name they should save that like jesus yeah yeah and also the early image generation stuff that they were working on with dolly that later became um they actually just yesterday announced it's been rebranded yet again it's now um gpt image yeah yeah so they went from dolly one to dolly two they have done an amazing job to compete with nano banana but we'll talk about that later in the show.

5:24So what was your reaction when you saw this early on? I thought it was terrible for a lot of reasons. And I argued that it was not the product they should release because it's really, these models are trained on documents, on document completion. And so that's their natural mode is to complete a document. And the task of sort of tricking that into being in a chat interface is very complicated. And it's led to a bunch of other really interesting problems. But basically, they're better at completing documents, which is why with agentic frameworks and stuff today, we're seeing this return to like, here's a JSON document.

6:04This part is missing. What should go there? And it's really good at that versus this chat interface where it wasn't obvious because back then, the context windows were really limited. So it may have forgotten the first half of the conversation and the user doesn't see that. So for all kinds of reasons, I felt like it was a terrible interface, but it's obviously the one that went viral. So here we are. It's what you know, huh, CJ? At that early stage, what were your recommendations, I guess, as an early product reviewer to the team? And do you get any sort of feedback internally on these sort of things, or is it just speaking into a void?

6:43So like one example was with Dolly. if you at the time it would give you eight images for any any prompt that you put in instead of now it's just one right um but the idea was like it's most of them are not going to be very good so we want a lot of feedback right so then that became part of the training set um but at the time there was this issue with inclusivity and like it didn't have any any awareness of the social context of the user and so like you know if uh if you ask it to draw your picture of an astronaut it's going to draw eight white guys right um and so it's like well this is obviously a problem like it needs to have some awareness of like who is asking the question what is their social context and instead they're like that's too expensive we don't want we're not going to retrain the whole thing from scratch we're just going to insert the word diverse into the prompt if they're asking about a person so instead it's going to change it invisibly to draw me a diverse astronaut and so then you get eight black girl astronauts and it's like this is even worse like this is even further from understanding the social context of the user and what they're really asking for and they're like well we're just going to ship it we're already behind stable diffusion at this point um and and so then like you know google adopted the same solution and then later you get like thomas jefferson is like a is black with a you know a native american war bonnet on and it's like, well, okay, now it's a problem.

8:08We have to fix it suddenly years later. But we were talking about a lot of these problems, you know, internally back then. And they're like, we don't care. It's too expensive. So, I mean, you'd have to agree they've, they have iterated and I don't, you know, I mean, you could iterate in private, but I think especially for products like this, it's probably better to iterate in public and take the, take the licks at the beginning in order to get a better product in the end and let us be the the beta testers sustainability though is still an issue and and you know we're really starting to see now i'm a i should explain probably that i'm a big booster i love ai i use it all the time i use it somewhat guiltily because we understand that there are a number of consequences the prices of ram chips have doubled uh maybe that's just a temporary thing the cost of electricity seems to be going up nationwide by orders of magnitude almost uh we just saw that you can't get skilled construction workers anymore they're all tied up building data centers or and then there's also the the financial consequences which i you know maybe or maybe not are something to worry about where pretty much all of our gdp in the united states has come from these magnificent seven ai companies that's almost all of our growth uh and then there's also the potential down the road consequence maybe we're already seeing it of job loss to ai uh i'm not a doomer but there are also those who say and once it gets super intelligent say goodbye to the human race so there are definitely a number of sustainability issues with ai but i notice you're not one of those people says, well, let's just not use AI.

9:55Yeah. Yeah. Yeah. Well, I would say, first of all, most of these concerns, um, I think are, uh, like for example, the example of electricity consumption, all AI consumes like less than half a percent of our electricity. Um, all data centers consume less than 4 % of our electricity. Um, so there were a bunch of forecasts years ago when exponential growth really took off with data structure investment. And it's projecting, you know, like 30, 40 % compound annual growth rate on energy consumption, on AI inference workloads and budgets. But that didn't happen. Very quickly, those compound annual growth rate forecasts were cut by 75 % or more.

10:41So like the current forecast is that we could see global data center energy consumption reaching three percent do you hear that sounds like breakfast is ready because quakers coming in hot with morning nutrition 100 whole grain oats and a good source of fiber to fuel the rhythm of your morning and kickstart your day and that sounds absolutely delicious. Fuel to start whatever's next. Quaker, official sponsor of FIFA World Cup 26.

11:20Ryan Reynolds here from Mint Mobile, with a message for everyone paying big wireless way too much. Please, for the love of everything good in this world, stop. With Mint, you can get premium wireless for just$15 a month. Of course, if you enjoy overpaying, no judgments, but that's weird. Okay, one judgment. Anyway, give it a try at mintmobile.com slash switch. Upfront payment of$45 for three month plan equivalent to$15 per month required intro rate first three months only then full price plan options available taxes and fees extra see full terms at mintmobile.com by 2030 of total energy consumption.

11:53So these are these are like trivial consumption levels. The reason that electrical energy prices are going up is because of a number of, like there have been a number of executive orders preventing completed renewable production facilities from being connected to the grid and cutting funding by tens of billions. Some of the analyses that are out show it's like more than$50 billion has been cut from renewable energy projects, including continuing to operate them. So there has been a reduction. There was a recent study that found over 113 million homes worth of electricity has been cut deliberately through these efforts to end the funding for renewable energy.

12:44And also there's the impact of tariffs on maintenance costs for the grid and for the sources of energy that we already have online and inflation at the same time too, right? So all the costs - These sound like political issues more than AI issues. And deliberate. Yeah. And so those are enormous impacts. And the reality of how much energy AI is consuming is it's less than a percent. So it's just not true. There are, I think, one of the things that's been most frustrating talking about AI ethics today is that the academic context focuses primarily on things other than generative AI, because no one has ever figured out how to make money with generative AI.

13:23None of these companies is making money. And the vast majority of the AI that's actually happening in businesses is not related to generative AI. It's insurance companies, it's decisions about whether or not a particular test should be given to a patient if it's too expensive versus like, what's the value of a person's life in terms of how many, you know, unnecessary tests we're paying for. Those kinds of decisions are enormously, you know, impactful in people's lives. And there are all kinds of ethical concerns around the construction of truth, the epistemic power that we're giving to companies like OpenAI that choose to put people like Larry Summers on the board.

14:03And, you know, the models that they're building are embodying that perspective of like these, these kinds of people. And is that who you want to be deciding what truth is for the world. No. Like, these are, I think, the bigger issues. And it's one of the things that's been kind of frustrating is seeing, you know, there's so much popular interest in, like, it's taking all the water, it's taking all the electricity. It's just not true, right? But there are also, like, alternatives, like these much smaller models that are performing better than the large data center models now. And they'll easily run on your cell phone on a free open source app.

14:37and you can get better results that way than you can get from these large labs, right? So I think the sustainability picture for me looks like smaller devices that run at the edge that run on solar power. There's this really cool, like, for example, have you guys seen the WizPlay open source project? No. What is that? It's a Raspberry Pi inside with a screen and a little face and you can hold the button and talk to it. It's like push to talk like a walkie-talkie. and on the device it will decode the text of like what you've said from the from the sound and then run it through whatever llm you want locally on the device and then locally on a raspberry pi yeah and then convert it into speech and it'll talk back to you and say the answer and it all is it good on this it's great it's so you can so this is using quen 3 which is you know yeah yeah it's at the frontier yeah it's great and and this will 16 gigs of ram you can run up to like a 16 billion parameter model or more if you miss a lower quantization yeah that's very encouraging really it sounds like uh and partly this is because of people worrying about AI power consumption and all these other issues that they're it also feels like I'm surprised it's too bad about renewables because it seems like the need for power would drive renewables as it certainly has in China for instance absolutely yeah we're going in the wrong direction here what about the we are commissioning a lot more nuclear power is that you consider that a renewable or a so china is currently building 150 new nuclear power plants the u.s uh is not right we're cutting them we're shutting them down right like you're shutting them down germany at least the safest cleanest most efficient source of energy that has ever been developed is nuclear power and we're shutting it down while other countries are building more and the marginal cost of electricity is going up here and it's going down in china right and so if the fundamental constraint of the ability to do artificial intelligence inference is how much power you can allocate to that um who's going to win right like it's it's no contest where do you stand on the on the um issue of a a bust a uh that we are in a bubble and that the bubble is going to pop is it going to pop uh yes i mean obviously yeah i would say yes 100 obviously how how to how to frame the answer like you you mentioned um right like if you if you look at oracle for example and its financial performance this year um there was this this deal made this company that was literally created by the CIA's venture capital arm is saying that they're going to build this enormous amount of new debt funded infrastructure for open AI, which doesn't have the money to pay them for that.

17:33Right. And, um, they are already levered at like an insane level, like way higher than the rest of the industry. And they're like, we're taking on trillions more. Right. And the stock jumped and And then it lost half and it's now down by half, right? In just a few months. And so I think when you talk about like, you mentioned earlier that all of our, more than 100 % of our GDP growth in the last year is the Mag7. And all of that is purely speculative and none of that has any ROI. And it's not clear that any of them even has a plan to get to an ROI. And so at some point the question becomes, how long are we going to, you know, keep these speculative bubbles going, right?

18:15The problem is people who have bought in can't sell because there's no one to sell it to without taking a 90 % markdown. And that's called a liquidity crisis. It's not called, this isn't worth what we paid for it, right? And so, you know, there are all these people talking today about the government should just buy all of these things from people that invested at these insane levels and bail out the investors. But, you know, even if, And whether or not that happens, you know, it's it seems pretty clear that we've been in something like stagflation for a while. We have high inflation. We have high unemployment.

18:48You have like, for example, the Ludwig Institute, which is this very well-reputated institute that does a ton of labor research and cost of living research has has been putting out numbers for years now talking about. We actually have like closer to a 25 % rate of unemployment. If you look at the number of people who are unable to afford the cost of living with the job that they actually have, you know, and the, there are all these other like alternate measures that people are doing independently of these things like inflation and like consumer price increases and, and, you know, people who have to work three jobs to, to pay rent.

19:28Right. um like we don't have low unemployment right now you know we have high unemployment right now and when you when you look at those factors together um you get the 70s right people are talking like andrew sorkin's got this new book um people are talking about it's going to be 1929 i'm not sure that's true yeah i think that's the wrong way to think about it so in in terms of like specifically the answer to your question is there going to be a big crash maybe not maybe it looks more like the 70s maybe it's a very long period of low or no growth or negative growth um and then you have i think that was already true before the the trade war um but you know it's gonna i think it's gonna be so much worse now right and you could see a 70s you could see a lost decade or more where we have like long protracted periods of of low or negative growth you did your um nba thesis on nvidia right yeah um i'm curious to that was how long ago did you write it uh it was two years ago right okay i'm curious about your perspectives on this start with chat gpt um i presume you think it's better now is there a point where you think okay they didn't listen to me at the beginning but now they're doing better Is there a point with NVIDIA?

20:52Do you think that they're doing a good strategy or on the wrong path? I'm curious, just your views of some of these big companies and where they're headed. So I think when you look at Chai Chippit today, one of the big bets that was made early on is this idea that if you increase the size, you will increase the capability by more than that. There's this like inverse, this exponential relationship between the scale and the capability. And that was true at first. But as everyone in intro to computer science learns, all exponential growth curves are just as curves you haven't found the top of yet.

21:32And we found the top, right? And they've started stagnating. They've stopped getting better. And now they're fighting over who's a little bit higher on this plateau that they're on, right? So I think the right way to think about it is what's called the platonic representation hypothesis. If we are all looking at the same object from slightly different angles, we might like draw in a slightly different way, but it's the same thing that we're all trying to converge on. And that's how all LLMs work. All LLMs are trying to converge on some platonic ideal in this high dimensional space. There's this shape that you can't even imagine.

22:09And they're all trying to be that. And they're all trying to get closer to that. And so oftentimes it's true that on many important benchmarks, you could see a 2 trillion parameter model performing worse than a 1 billion parameter model, right? This is what Andre Carpathy was talking about in his really good recent interview with Dorkesh Patel about how like the future is probably like these 1 billion or smaller parameter models that know how to solve problems, how to find the information that they need and put it together. And so that's what I've been arguing for a long time is that like smaller models are better.

22:43We should have models that we can, you know, stick on a Jetson and run at home and not need to pay anybody rent for that. Right. That's just small, but open also, right? Yeah. And we have that, right? And so one of the things that's happened recently is opening, I actually changed their strategy. They have abandoned their religious reliance on the scaling laws. And the new version has a built-in router that will send you to a small model if it feels like that's enough. And you don't get the big thinking harder model unless it feels like the small one isn't going to be able to handle the question.

23:19So I think a lot of these things are already being adopted in an effort to do cost cutting and just the realization that bigger is not always better for every situation. And in terms of NVIDIA, my argument was that Jensen has been preparing for a large reduction in demand in the near future. And so you had, for a short time, there were companies like Microsoft and Facebook buying hundreds of thousands of H100s per year, as many as they could get their hands on all these other companies couldn't which is why like for example this this um dojo supercomputer that's using like um feature sizes from the iphone 11 to build these like the worst ai chips anyone's ever seen um and the person in charge of it quit they hired a new one he quit they hired a new one he quit and then they just shut the whole thing down so tesla effectively has no ai infrastructure at this point and they're renting it from this new xai startup um which is chinese by the way oh no xai that's grok yeah i'm sorry i thought you said za i was thinking za i so confusing with the letters so my argument was like nvidia controls 98 of the market at that point for um inference hardware uh and they charge an arbitrary thousand percent markup and everyone is paying as for as many as they can buy right they should be charging more they should charge a lot more and they should buy a platform like hugging face that gives them the power to make it free for developers and um let them rent the stuff out instead of uh selling it right um and so they did that actually a month after i wrote the paper with this core weave not because i said it i'm like i don't think it's i don't think anybody read it take the credit but in the paper i was like this is so stupid obvious that it must already be the plan and then it is what happened so that's how we got car weave essentially um so today that you know there are questions about like this whole michael burry thing with depreciation um it's like a big topic right now and it solves that problem because nvidia is renting them out they they aren't charging themselves a thousand percent markup right so they have like much more margin there to work with in terms of depreciating over a longer period and giving like whatever prices they want to the market if they control the entire inference market and not just the market for hardware that inference is done on.

25:53And so, you know, like historically, I think they're like less, they're less evil, right? Like NVIDIA is not like an evil company in general, where you see like a lot of ethical concerns with the way that inference is used, not so much with the way that like the hardware for it is being built. So I think that that ethical landscape is more interesting in terms of like what is allowed to happen on these platforms versus like who's who's buying them. We're talking about the Dematrons. CJ Trowbridge, who is you'll see him on TikTok, them on TikTok, Twitch, YouTube, but is also an expert on sustainability and has a doctorate in ethical AI.

26:38It's not a doctorate. Not yet. Not a doctorate? No. A master's in ethical AI. A postgraduate degree. Am I right? Is that right? Is that fair to say? I'll have to look at your CV again. Yeah, so I did my MBA. And then instead of electives, I did the ethical AI program, which is a graduate certificate program. Ah, got it. At San Francisco State. Tell me about ethical AI. It's almost an oxymoron. What does that mean? uh well ethical is the science of deciding the right way to do something and the wrong way to do something um and then um what is the right way and the wrong way to do ai so historically it was primarily concerned with issues around like there's this this idea of like receiver operating characteristics there's some relationship between two variables that is like ideal in terms of the ethics.

27:32So that might be how many expensive tests are we willing to pay for in exchange for how many people are we willing to let die because we accidentally said they don't need the test for an insurance company. So what is the value of a human life compared to what is the value of the test that we have to pay for? So historically, this is one of the big ethical problems in AI, And today as well, right? But we have these other new emerging ethical concerns in AI, particularly around epistemic power and who has epistemic power, which is the power to tell you what is true, right? And what are the perspectives that are being embodied in the epistemic power that's being expressed by a company like OpenAI that's choosing to put the kinds of people who it is on the board, right?

28:20And someone like Sam Altman, right? um so are these the people that you want to be in charge of telling the world what is true and false um and you know if if yes why if no why right and those are like the interesting ethical issues i see in generative artificial intelligence that i think don't get enough attention because people are so focused instead on concerns around um resource consumption that is is is less than a percent of our consumption. This is not the issue. Why do you think that that issue has become so sticky? I assume you must experience this a lot given the size of your following on TikTok, but why do you think that people have kind of glommed on to AI is taking all my water and not any of the thorny or more pressing or existentially concerning issues you just detailed?

29:09Yeah. Well, so let me give you an example. um the the most salient version of that argument to me that i have seen is people who are arguing that xai's data center in memphis is is taking so much water that it threatens the ability of people there to have drinking water and it's like okay um i i am like talking about elon has paid my bills for years right ours too ours too as someone that has met him. I met him at Burning Man. We have a lot of mutual friends. He had no idea who I was, which I'm like, I'm so glad. Thank goodness. Don't have me killed, please. But what's the defining geographic feature of Memphis, right?

29:53The amount of water that is going through the Mississippi River in Memphis every second is 300 times as much as that data center uses in a year. And the data center is built next to a sewage treatment plant. It's using treated sewage for cooling. It's not using drinking water. Like it's just every part of the argument is just, you know, no one looked anything up. But at the same time, like there are these real concerns around like air quality, right? They have 30 gas turbines running in the parking lot to power the building because they didn't bother to get approval to connect it to the grid. And it's the perfect place to build the data center because the TVA, the whole purpose of the TVA, the Tennessee Valley Authority for Electricity, is that it's this New Deal project of like, we're going to build power dams, we're going to build lakes, we're going to produce an enormous amount of abundant and cheap energy with the hope that we can entice aluminum manufacturing into the area, which never happened.

30:54So they have all this excess capacity, they have all these nuclear power plants there. It's the perfect place to do that. But they never got the permits. It's like they never, they never, they didn't check where is that power available. They just picked a random spot and built it there. And it's like, there's not even power lines, you know? So he brought in all these emergency backup generators, which are illegal to run that way. And it causes enormous pollution, which harms the surrounding neighborhoods, which are primarily black neighborhoods. And he's talking now about how he's going to build a two gigawatt natural gas power plant there.

31:33And it's like, you can't just do that. Like, people are going to die, right? So it's super dangerous. These kinds of other externalities, and these things have nothing to do with AI and everything to do with the unethical and incompetent activities of people like Elon Musk. I'm so glad you're saying this. This is such important information. Yeah. And it's frustrating to me to see people saying, like, AI is bad. It's like, that is not the lesson to take from this. Elon is bad. Elon is killing people. AI will run on your phone all day and consume no energy or water. AI brought this video to you on your TikTok feed saying AI is bad.

32:17Yeah. I think there are bad ways to do AI. Absolutely. Who are some good guys in AI? I mean, I think everything that China is doing has been amazing. in AI. All of these free models that are more performant than the best stuff you can rent in the cloud here, that you can get better results, that it was trained on these nuclear power plants, it's not being trained on coal and all the terrible energy sources we're using here. It's being trained on these much cleaner and safer energy sources. It's being trained on... There are all these concerns about like the the the con like what what it what is the ownership of what's being used to train it and we know that in america we have companies like anthropic paying out more than 10 percent of their last funding round in lawsuits because they they torrented all these books they stole all these books from authors they didn't pay for them they didn't use things that were open source they didn't use the public internet they just started torrenting stuff and stealing things from people in order to train Claude.

33:25And so they paid out billions of dollars in settlements to authors. And you have the same thing happening at Facebook. They were also doing that. They were using Anna's archive, this like huge archive of, you know, torrents that they used to train the Lama models. And one of like, you know, there are superior training sets available now. Like FineWeb is an open source training set. It's 15 trillion tokens. That's more than enough to train NLLM. And it's what was used for, like, Andre Carpathy, after he quit OpenAI, published a CUDA library that he wrote with instructions and bash scripts that would let you replicate all of OpenAI's products yourself using superior training sets like FineWeb that weren't available when they trained those models.

34:19So if you did, if you train GPT-3 in the same way that they did originally on these new training sets, it's actually significantly better than it was. So like, there's no reason to be stealing all this stuff. Like the stuff that's in the public domain that's open source is actually superior and it's free. If they had bought one copy of every book, they'd be a lot better off than where they are now. Yeah. Yeah. So, wow, there's so much I want to ask you about. Yeah, your thesis in ethical intelligence was about metaphors and how we misuse metaphors. We call it intelligence. We anthropomorphize it.

34:57And that really leads us down the wrong trail. I've been talking about that for some time. But it's hard to find another language, especially a language people understand, because we've all kind of settled on this idea that it's human-like intelligence. Right. Do you have a better way of talking about it we can use? I think it can sort of relate to what we were just talking about. I think it's frequently the case that engineers will be assigned a task like, there's this criticism out there. It doesn't have reasoning. And so the engineer sits down and they're starting to type a function and the function definition, what's the name of the function?

35:30Reasoning. Okay, how am I going to do that? And they'll invent something totally unrelated to reasoning. And it's like, okay, now it's solved. It does reasoning. Right. And this is true with so many of these metaphors where you have engineering teams tasked with solving fundamental limitations of these tools and these products, and they're calling it that, and it's just not that. Thinking was the same thing. Reasoning is the same thing. Attention is the same thing. What we call attention is not at all related to what attention is. It's just the fact that they have these multiple hidden layers that are fully connected.

36:02That's not attention. It's not making any kind of executive decision to focus on this or focus on that at all right yeah and in the same way it's like the variables and the functions yeah think of it this way it's really amazing how concrete that is yeah right and you can see the same way with like we need training data we don't have training data let's steal it right it's not like let's curate a an ethically sourced open source library that we can use to train it it's like no no no let's just do crime that's way easier it's faster it's easier let's just this is that silicon valley uh yeah think of you know move fast and break things and you know apologize later well and like eric schmidt in his talk at stanford it's like everybody should just go out and start ripping off everyone else's ip with vibe coding and if it works you'll have the money to to fight them in court and if it doesn't work you just close it down and walk away nothing nothing than to sue for yeah you say you say and i think you're right that the models and maybe it's because of this become commoditized that there isn't really that much distinction between any of the models right because again i mean there is this this platonic ideal shape right if you think of like parameters a parameter in an ai model is essentially a connection between two neurons right so you have um billions or trillions of parameters in an ai model and it's just a list of numbers that you're multiplying there's not more than that that's going on you have this really long list of numbers you're doing matrix multiplication it gives you the the next word right um that is those parameters are dimensions and so you can think of these hyper objects uh as like a two trillion dimensional shape like we live in three dimensions what does two trillion dimensions look like you can't even imagine that right but that's what they're all made of and And there is some ideal shape that they're all trying to converge on.

37:59And so one of the things that's really interesting is the failure of the scaling laws is like you have benchmarks today where there are 2 trillion parameter models that will score worse than 2 or 3 billion parameter models. And so it's like, what is the other like 99.999 % of the stuff in there doing, right? it's not helping it's not necessary there's this enormous latent space and it's because we're using unsupervised learning um you know the the process of training is like here's a whole bunch of stuff read through it until you feel like you can predict the next thing wherever i start you right and that's how we train them and then it's like okay how can we make this be a chatbot now and so like a lot of these things are like this just isn't their natural mode and um the thing that they're actually trained to be good at is not what we're actually like using them to do um and it's part of why there is so much this enormous latent space that is not like useful for actually performing the tasks that it's being used for us because it's like this unsupervised It's sort of like figure it out, good luck sort of training method.

Read the full transcript

39:15My sense is, and I think you're, I hope I'm phrasing it properly, but I think you're right. The issue isn't AI. The issue is the people who are bringing us AI. Absolutely. And the choices they are making. 100%. For us. So let's do it better on our own. Let's do grassroots stuff. Where can I get one of those little whisper things, whatever that thing is? You can't. I want that. so it's it's it's an open source project uh it's called with play and essentially it's this is like a 3d printed case it's got a raspberry pi with this little hat that has the screen and the microphone and the speaker and the button and then you can print a case for it and you can just sit here and talk to it oh you can buy the wisp on amazon os is currently available yeah yeah it's a Raspberry Pi Zero?

40:07No. No. It's a little more than that. This is the 5. There's also a bunch of these... Pisugar.com allegedly has it. That's the manufacturer, yeah. So, like, you'll notice that the hat is the same size as... It's the same form factor as the Pi Zero, right? Yeah. But the Pi Zero only has half a gig of RAM, so, like, it's not running any LLMs. But there are all these new, like, Chinese clones like this is the um orange pie which is the same form factor but it has four gigs of ram but it doesn't have matrix multiplication extensions for the cpu so it's it's gonna be like even slower there's some there's a new one that they haven't actually started manufacturing yet from radx or radsa called the qba7z that's got 16 gigs of ram with an npu in the same form factor so you can do you'll be able to do like with half the physical size and even even more powerful models which is really exciting we're in this really interesting place where there's this like abundance of like small edge stuff that's getting really good so it's like it's really exciting this is exciting if somebody uh and i know probably most of the people listening want to start doing this on their own what do you what do you recommend do you i went out and bought a framework desktop uh so that i could run local models downloading them from lm studio at least that gave me a sense of what was out there.

41:34Where do you start? What do you do? LM Studio is great. It's a free open source project that's built on top of the GGML library. GGML library is incredible. What it's accomplished is this open source project to make it really easy to run AI models for inference without, you don't need TensorFlow, you don't need any of that stuff. It's just this very simple library that can be included in apps. There's all kinds of apps like Ollama, llama cpp lm studio um that all are built on top of that olama is i mean it's a meta model though do we care where the model comes from no so olama is a server that gives you the same api that open ai uses with your own models so you provide your you don't have to use llama with olama okay okay no yeah yeah in fact that's olama is what everybody uses for all of their local models That seems to be the kind of dominant Lama CCP seems to be the Lama CCP.

42:30PP. So, okay. So, and then what models - If you wanted to do it on a cell phone, whether you have Android or iPhone, iPhones are really constrained with their memory resources. There's no iPhone that has more than eight gigs of memory. So realistically, you're going to be way less limited on Android. but both of them have a free open source app that's also built on top of the ggml library called pocket pal and um it's an open source community project and it'll let you run any of these models on your phone and you can do it all day without needing to plug it in which tells you that like it's not taking all the power right so if you're running these smaller models or you're running them in smaller amounts of memory what are you giving up they're not as smart they're not as What do you lose?

43:20Some of them are smarter. I think the really interesting stuff that's going on right now in this community is including all of these same tool capabilities. So for OLAMA, there's a web front end you can use called OpenWebUI that will connect to OLAMA. And these are just Docker containers. You can run them with a single Docker command. And then you've got an interface like ChatGPT that has the same layout. You can just choose whichever model you want. You can download whatever model you want right there in the web UI. You don't have to get into the code at all or anything. And it includes tool integrations that are converging on all of these different paradigms that are being adopted by different labs.

44:03So it can already go search the web for you and bring back the answer. There's also all of these really interesting projects, including one that we're working on with um one of the non-profits that i started we're doing um like off-grid like solar powered systems that are is this cyber pony cyber pony express yeah yeah yeah the cyber pony express mesh tastic network yeah yeah it's off-grid yeah yeah and it's got ai on it so you can chat with ai on it so i think it's also i like the future that you're building here this future of local small ai's not run by big tech companies and kind of demented tech bros but created by the people for the people with the use of the people and it's small and it's nimble uh it feels like that's that's a great direction to go yeah so that that project we've we did a gofundme we raised like twelve thousand dollars to buy the radios and just give them away for free to people so like these little mesh radios yeah and so it has a range of like up to 20 miles to the next person and the mesh that it builds is huge so right now you can it's like nevada city to gilroy from like modesto to sonoma you can you can text on this network that there's no internet there's no cell phone towers um and it's all end-to-end encrypted no one's in charge of it no one could be like in control of it it's like a totally like anarchist peer-to-peer type system and we've been doing a lot like the team that's building this this software is um very into burning man so we did like this year at burning man we did a group chat with 10 000 this sounds like something that would happen at burning man all right now i get it yeah yeah yeah so i'm also the regional chair for burners without borders for the bay area so we're interested yeah we're interested in uh we do like disaster response disaster preparation love it love it and like there's this interesting sort of overlap of like the prepper community and like the ai community where it's sort of like can i use this in an emergency can i use can i use the mesh to like contact my friends and family during a disaster make sure they're okay and it's like yeah and it's all solar powered by these little devices that people have put on all the mountaintops and you can just send messages across them the next generation of amateur radio of of hams hams have been very active in local disaster prep but yeah you had to get a license to do it you had to like get some that's before yeah said you ruined radio but yeah this is this is great i love this well i yeah i mean i don't know i i am also a licensed amateur radio uh tech and i like the the thing about like those organizations that you're mentioning like aries for example there are all these amateur radio emergency organizations where people with an amateur radio will help during a disaster because the infrastructure is all there and it's not reliant on any of the internet or cell phone infrastructure um it's so like for example this year at burning man we had 10 000 people in a group chat on my chastic the whole city can talk can listen can get news from each other about what's going on and we had a number of disasters uh and we had like people with like there were meteorologists with like they brought a doppler radar and they're like okay here's what's about to happen and that's so much more accurate than what you can find online even right and so it's this these kinds of alternative like i love this peer-to-peer mesh topologies are just inherently superior to like legacy media institutions in terms of like their ability to disseminate important information hyper locally how do we get this to let's i want to let's get this going how do we get this going we are what is the high desert institute tell us about that uh yes i've uh a few years ago i started a non-profit with a group of friends and we basically buy uh land all over the country and build like free communities for people that want to go live there and build houses and off-grid yeah and we there's a tv show about the most recent one um on discovery which is kind of like they filmed the process of building it and then they like sort of tried to make a reality show out of it which is like it wasn't a reality show but yeah of course yeah but it's out there if people want to check it out um but basically we've been building these free lands all over the country and one of the requests that we got at the first one was someone was like why can we build our own internet so we don't have to use like the government satellites yes and uh at that point There wasn't a lot of stuff like this, but MeshTastic is still very much a beta project, but we're working really closely with the development team and building out these networks.

48:46Now it's really possible to have this secure offline communication with people. We've got a bunch of robotics projects, for example, that we're working on that's integrated with the Mesh so that you can put an LLM on... I've got like the quadruped like canine you know you can 3d print these now and well you put that on a dog or it is a dog you put it on a dog the legs are not attached but yeah you can you can have it be on the mesh and you can give it simple text instructions and it's got an lolm on there that'll interpret you know the context of where it is and what what's the instruction and figure out how to go do it this is much more the neuromancer future uh than it is uh the dystopian future others have imagined this is a i love this idea where you're going to have these hyper conglomerates is kiretsus you're going to have these big powerful companies yeah let's move under the radar at a right scale and make a difference and do it for ourselves so we're not stuck in that machinery yeah i wasn't sure where you were going with that because so like romancer is my favorite book i'm like that's pretty dystopian but it is dystopian and it's also about these giant corporations but there's also this underground right right if the future sucks because big corporations are evil how can people use technology in unexpected and unplanned for ways to be okay through that and survive and thrive through the chaos that you know we're likely to see in the next 20 30 years this is what we need to do yeah absolutely i'm excited and very nervous about the the tv show adaptation that's about to come out i am too yeah no please don't cjtrowbridge.com uh is probably a good place to start uh you can see all the projects they're up to and all the theses are online his spotify or their spotify playlists are online people ask for those it's it's all there uh but i think the thing to find out about is is uh all the all the different things CJ's into and their TikTok and YouTube and Twitch streams.

50:52And I just think this is, I want to support this because I think you're on the right track. Absolutely. And this is where we need to go. I think we've, don't you think Jeff and Paris, we've kind of come to that conclusion over the year doing this show, thinking about big techs, AIs and what they're doing and what's happening. And I just think it's very encouraging. We've heard more and more about small models over the last few months. I really think it's exciting what's about to happen. Yeah, absolutely. Yeah, I'm very inspired by it and encouraged by it. CJ Trowbridge, thank you so much for joining us.

51:30I would just add sort of like a last thought is that if the bubble pops, if all these companies go away, almost all the value of what's been created, we will still have in these open source models that we will continue to be able to run on the edge. And the reason we're not making a lot more of those is because we've kind of reached what we're going to reach. This is the end of the plateau. And so no matter what happens, we will still have these tools and we need to figure out how we can use them locally and not rent them from these future ex-corporations that are currently ruling the world. Love it.

52:08But so nice to meet you guys. Your lips to Elon Musk's, nevermind. Long gone. Thank you, CJ. I really appreciate your time. Very inspiring stuff. CJ Trowbridge, everybody. We'll be back with the AI news in just a bit. Our show today brought to you by the agency. Here's another project I think is pretty worthy. Build the future of multi-agent software with agency agntcy it's now an open source linux foundation project that's so cool agency is building the internet of agents it's a collaboration layer where ai agents can discover connect and work across any framework and the thing is all the pieces engineers need to deploy multi-agent systems now belong to everyone who builds on agency that includes robust identity and access management that's super important you've got it it ensures every agent is authenticated and trusted before you interact with it agency also provides open standardized tools for agent discovery seamless protocols for agent to agent communication modular components for scalable workflows you'll be working with developers from the best cisco dell technologies google cloud oracle red hat 75 plus other supporting companies to build next-gen ai infrastructure together agency is dropping code specs and services no strings attached visit agency.org to contribute that's agntcy.org we thank him so much for supporting intelligent machines what do you think jeff i think uh there is some hope yeah yeah but i don't think we can write off the big guys i don't think he's saying that either but i think that we gotta force we to push them still yeah we got to push meta to have llama and um i mean do you think llama's long for this world that's what i'm saying i don't think it is that's what that's what i think we've got to keep the pressure on and not just give up on on trying what would keeping the pressure on even look like at this point this talking about it you know i guess i think jan la quinn leaving well matters anyway i i'm i don't know whose idea was i think it was anthony's to talk to cj but i'm super glad uh we did i think yeah what a great uh yeah that's that's yeah if if there's to be a future for for your generation paris uh it isn't i i'm assuming but i'm acting as if there's not just you know that would be hey listen there's still i mean what's interesting is the big guys are still doing interesting stuff this week two new models out they're competing with each other like crazy gemini 3 flash which is a much smaller model built for speed not open source it's gemini look there's uh there's our friend uh tulsi dolci writing about this do she writing about this on behalf of the gemini team and the keyword blog uh people have already tried it think it's very impressive it's very fast thinking even thinking and its scores are very good uh it's also cheap relatively cheap gemini 3 pro which uh is two bucks for uh a hundred thousand tokens or i'm sorry price per million tokens is two bucks price per million of gemini 3 flash 50 cents this is a quarter of the cost call now call now now it is more expensive than 2.5 flash but the flashes have always been a little bit less expensive did third did well uh so gemini 3 pro 37 on humanity's last exam um flash 33.7 so close very close in the scores and you can see the other scores here in fact it did better on arc agi 2 so the this is a this is uh they're moving forward not to be outdone by Nano Banano ChatGPT Images is here and I've seen a number of comparison between the new images in ChatGPT this is, you know, their old dolly but they've updated it and the image is created by Nano Banano and I have to say it's not only competitive I think in many cases it's more realistic it feels more real yeah let me see, I'll find some examples on Reddit there have been some good examples.

56:43I was about to say, I feel like every example I've seen on Reddit has been the opposite. Oh, really? Mm-hmm. Okay, well, today Jason tried out - But it might not have been the most recent update. It's also prompting, you know, and so Jason tried it? He tried out Gemini versus the new ChatGPT, and Gemini was beating the hell out of it. It was? Yeah, better and faster. Images, you mean? Well, he was doing, what do you call them? Infographics. Oh, yeah. The infographics are incredible in Nano Banana. Incredible. I saw a family, a candid family photo comparison between Nano Banana and the new GPT images that I thought GPT images was indistinguishable from.

57:32I'm posting a haunting image on in. I guess it's I guess, you know what? It's it's again, it's just it's hit or miss, right? Right. And it's leapfrog. Right. People have also, I mean, as per usual, been really mad on the chat to be open AI subreddits that they're worried it's taking away their horny mommy. For lack of a better description. I'm sorry. What is that? By which I mean, people are upset that either their AI girlfriend, boyfriend, oh okay partner or their comfort object that uh perhaps takes a parental figure is gone they're upset that this iteration latest iteration of chat gpt that we've had come out last is not your not your horny mommy anymore for lack of a better description yes i thought it was pretty what the kids call it no i don't think i've ever heard of it but i do is that you did you make that That's fresh off the dome, baby.

58:36It clearly is not confusing to anyone at all and didn't require a immediate explanation. No. Yeah, I'm looking through some of these images. Well, you know, here's one I saw. This is an example of making a World Wildlife Fund ad, One Planet Many Lives. the prompt uh in both cases was use a globe on moss images the main background inside the glass globe at a small panda pair realistic softly reflected curve pair p-a-i-r not p-e-a-r curve to match the sphere tag one planet many lives small line optional protect habit of drastically i think they're both very good i kind of like the image to jet gpt image a little bit better yeah no gpt no i like nano banana better you like nano banana better okay so that's a matter of preference I guess nano banana is better at topography it is better at typography by far i agree with you i thought you said topography like topography or typography or typography yeah i'm just shows where my brain goes to the plan In any event, even the big boys are improving rapidly, finding new ways to do stuff.

59:58But I kind of agree with CJ's thesis that they are a little bit commoditized. Oh, yeah, absolutely. Yeah, they're very similar. This is also where the different paradigms work. There's a paper today, Jan LeCun is one of the co-authors, and I had to have AI explain it to me, but deals less in words and more in ideas. And only later in the process does it translate the idea into words. right this is kind of where this stuff's going it's i feel a little bit like a like a junkie though using this i can't stop using this stuff even knowing if it's you know i'm glad no it's not bad for the environment but knowing that it's not good for humans what do you mean you can't stop using it i i'm hooked like claude what have you used it for today okay claude code is today i just use it now to i was having a weird problem with my emac setup i had it i said take a look this what's going on and it solves it it's very good um i i do images all the time um i don't have a right things ever i don't think um i i mostly mostly i would say the number one use is research so if i like just as you did jeff explain it to me um i ask it uh you know medical conditions.

1:01:14I ask it for advice on, you know, what's the best way to fix a squeaky door, that kind of thing. It's very good. I mean, look, it hallucinates. I had a problem with my Linux setup, and I was not, I tried and tried and tried with all the different models, never was able to get a solution that actually worked. So yeah, I'm not saying it's perfect. In fact, we have a story here from one of the best Mac bloggers about not using AI to solve your Mac issues. He tried it against a couple of well-known. I also say I don't think you should be using AI to solve medical issues. Not solve medical issues. Or I would argue you should not rely on AI for medical issues.

1:02:08um yeah i mean you have to be judicious i'm not saying you shouldn't be judicious but i think part of the issue is that these systems are designed to make you not judicious about it because it seems confident and yeah well i know more i know better than that and you're right i think that's a pitfall but i think no one questions to ask the actual careful about that yeah but i've but i found it uh incredibly useful um this is this is an example this is from uh the eclectic light company macintosh blog last week on my mac how good is ai at solving mac problems he gave it three i think fairly common issues uh with macintoshes why is it so expensive he says google's ai can't maintain a logical sequence and troubleshooting prefers to direct the user to command tools rather than the familially bundled utilities and all the answers were wrong how to reset permissions they were just wrong now he's an expert uh so it's good he knew they were ever that wrong when it's talking about your precious medical issues though well but it isn't oh you're so mean to me why are you so mean uh no it isn't i you know and but that's the that's the caution is if you didn't know and you typed in the commands that the ai gave you on the mac you might screw your mac up um me it's just my body it's no big deal yeah time magazine names ai the person of the year thing it but it's not even and then do you guys know the poly market aspect of this yes that's hysterical so people tell us what's that let's first let's show the time magazine cover okay there's there's two there's one okay zoom First one is just something that says AI.

1:03:54Yeah, when they're building it. But here's the so-called architects. Skyscraper. It's a takeoff on a very famous 1930s picture of construction workers over New York City sitting on a girder eating their lunch. Instead, it is, I don't know, eight people who are somewhat involved with AI. You've got Mark Zuckerberg. You've got Lisa Su of AMD. I don't know why she's there. you've got elon musk some women in the president they want to put some women in they've got jensen wong they've got sam altman there's demis hasibis uh that is dario mode of anthropic and then weirdly they do have another woman feifei lee who is probably one of the along with demis demis probably one of the two real architects yes that people actually do anything but poor feifei lee is sliding off the girder halfway even though there's plenty of space to look at all mark But Mark won't move over.

1:04:51By the way, Mark has never looked so happy in his life. That is not. Anyway. So I put up a cover. So both of these are the time cover of the year. Ostensibly, it's for AI architects. And listen, this is going to be the only time I'll ever come to the defense, and I'm not even doing that, of polymarket betting people. And I assume it's polymarket bet. It might have been the other one. Kashi? Kalshi? I think it was both. It was both Kalshi and Polymarket. In both cases, they had some sort of bet going on for whether the time person of the year would be AI. And when these covers went up, people were like, hey, yeah, I voted yes, that it was going to be AI.

1:05:31Where's my money? And in both cases, they closed down being like, sorry. You lose. AI wasn't the person of the year. Architects of AI. Well, there's also, and we were talking about this on Sunday on Twit, and it's interesting because we had a good panel. One of the people on the panel, Owen Thomas, worked at Time Magazine during some of these debates over the cover. And he said, these guys completely punted. They should have picked probably Jensen Wong, but they just couldn't pick one. so they these covers are such bs it's just to sell meaningless it's meaningless the discord yeah jürgen schmidhuber's version oh do you remember that one year that the person of the year was you was you you uh they did the same thing they did the same thing with the personal computer they wanted to make it steve jobs but he was such a jerk they didn't they didn't dare do it so they decided to make it the personal computer here is as you were saying some of these aren't architects in fact two people are but here's another architect schmidt humer's version of the person of the year who is that yeah we've talked about him jürgen schmidt humer is one of the uh ai has many parents jürgen's one of them and he's apparently very strong it looked like it i think look it's clear that none of these people except for again demis uh possibly dario and for sure feifei lee have sat down at a computer to code any ai um but uh so architects maybe i guess the architects could be the financial guys is jensen a coder i don't know joel jensen's technical he's more technical anyway yeah he's more of an engineer but yeah it's just it's just it's just yeah we shouldn't get too excited about that because it is after all just a marketing stunt i told the story i think when when when my my mentor people magazine pat ryan was facing the cover with mel gibson had nothing had no idea what to say and she said oh screw it let's just call him the most the sexiest man live and that became and that started the the franchise the trope yeah yeah so it's just it's just it's like 10 best bagels in new york we were talking about this on twitter and and the irony is it used to be to sell magazines right you put it on the cover it sells magazines but there's no newsstands there's no magazines what do you put what are you doing now why do you do this is it to get link clicks i maybe it's to get us to talk about people angrily tweeting you and we did exactly and uh and we got some people upset at polymarket and calci for sure um

1:08:18by the way the claude skills that we've been using very effectively in our research for the shows thank you anthony nielsen have now moved to open ai as well it's such a open ai is that a skill Was it really used for that? Yeah, well, it wasn't initially, but once Claude added skills, we decided to - I need to figure out, I guess, how to turn that into a skill for mine. Yeah, well ask it. I mostly just have it in a project. It'll tell you how to turn it into a skill. I guess it will. So now, that was back in October. Now, here we are three months later, my God, OpenAI is doing the same with ChatGPT and their command line uh codex everybody's got it is commodified yes and and by the way cj's point was it's all trained in the same data so of course it's commodified i guess that makes sense it's all transformers and it's all attention must be paid and it's all not on currently unprofitable and it's all unprofitable although there are some here's a company that says it's making money 100 million a year it's chat gps is that revenue or is that profit yeah revenue yeah revenue this is although i think if anything's going to succeed this is probably the right idea open evidence has a uh ai for doctors only it's uh and it's trusted unlike the ai i use for my medical advice it's trusted because it's based entirely on medical journals on peer-reviewed medical journals like the Journal of American Medical Association and Lancet and so forth.

1:09:56They have now a valuation of$12 billion, according to Stephanie Palazzolo at The Information, on revenue generating annualized revenue of more than$100 million. And how do they make the money? They sell ads to pharmaceutical companies on the advice they give to doctors. Oh, boy. so there is money to be made you see it's gonna ruin ai too yeah yeah i i'm presuming they're not putting it in the actual answer that they're you know why would you ever presume that i don't i don't know you're right good question you're being too kind of yeah i i can't i can't use it you have to be a physician you're gonna have to click through three to five pop-up ads It's about brand name drugs before you can get your diagnosis for your specific stage of terminal cancer.

1:10:53That's going to be our future. And even if it does say sponsor and it's underneath the result, that's got to have some influence on a physician who's looking for an answer to a question. And he looks below and says, oh, yeah, I need to truceeba liba dba. That's what I need. I like their pens. There can be more pens. I want more truceeba liba dba pens and pens. They buy me a nice lunch whenever I'm able to actually open my email underneath all these pop-up. But that does happen today. Pharmaceutical companies are talking directly to doctors all the time. They hire attractive young women who go from doctor's office to doctor's office with swag bags.

1:11:27And it's, yeah, I know a number of people in pharmaceutical sales. Sam Quinones' great book about opioids. That's right. That happened. That's right. How did that, whatchamacallit, family succeed? Yep. so did you like the story i made for you yes i loved it isn't that cute i like the first one with paris as a little girl did you tell i don't know why i'm so short in your time either so very tiny gemini storybook is out and it'll make a you just give it a prompt you put a picture in it'll use it and it's is it for parents to make stories from their kids anymore for smart asses to make them for their podcast friends let me uh let me see if i can uh i can find one here oh you emailed them i emailed them oh they're so big they had to come yeah they're so big you're so big we can't so you can do them in anime and claymation and just plain i like the claymation one but that's the one where paris is really teensy weensy well it was like i said i think you're old and both of them two fathers it confused oh that's how that it confused us so i'm Leo and Leo is me.

1:12:39Which one would you like me to show first? I don't care. They're both. Operation Prawn Star? Yeah, I think so. I think so. Wait, did you send multiple? I only see one. Oh, there's two. Yeah. Well, let me show you the first. I'll show you both of them. Oh. I have to download them. They are large. Pharmaceutical swag is banned in California. Well, I didn't know that, Scooter X. Well, that's good news. That is good news. So here we are riding our boat, the podcast pig. There's Paris in front. That's one way to call it. Yeah, trademark glasses. I guess that's me. And then Jeff behind me with a mustache.

1:13:16Oh, no. But now he's got a goatee. Hello, Google Island, and welcome to a very special broadcast, shouted Leo, leaning into an enormous foam-covered microphone. This is Paris and her two dads. And today we are embarking on Operation Prawn Star. Have you seen this before? You saw this one, right? I haven't seen this one. okay harris nodded seriously tears over the name podcast pig but adjusting her headset our mission to find and capture the elusive glowing and slightly aggressive radioactive shrimp it's rumored to live deep in the fizzing fjord beyond the bubble sea oh my anyway i won't read the whole thing to you this is claymation though all through the claymation there's so why do you look why did both of them make you look so young i know it pisses me off well i get older as it goes on i look it's because i love that it says shrimp guide on it that's very cute yeah my hair is not brown jeff you look like burl ives i don't know i the only way i know it's paris i've had every different hair color through us i know it's the glasses of the trademark don't panic dads paris click buttons on our keyboard control panel we can use the sound to our advantage high frequency bait and here's the radioactive shrimp but jeff's apparently interviewing i like that it's jeff is in a boat but it's also underwater yeah i don't get it um oh but it says leo raised his microphone so now i don't know who's who who's whom wow uh who's who next week we're countering the mountain of mildew i look like i've just seen a ghost in this panel this is good you want to see the other one i've seen you've seen the other one all right we should show the audience the other one it's cute yeah they haven't seen the other one so we'll uh so if you're an actual parent with a little kid this could be a lot of fun should i be sad you know uh patrick delahanty our um uh cto his wife is a very accomplished very well-known children's book author she illustrates it she's finishing up book it takes her a year none of these is going to be a bestseller ladies and gentlemen i hope not so this is also i mean i as the only person you hear doesn't have kids i am unqualified to speak in this one speak as if i'm qualified anyway because i'm a podcast because you're a podcast i would imagine if you're reading a children's book you're going to be reading that god knows how many times over the course of multiple years yeah you are going to want to purchase a book that someone's actually taken the time to illustrate and time to write because you're going to be looking at those pages god knows how many i've read good night moon thousands of times in fact we had to buy many editions of it because the first one we got abby would tear the pages out and eat them no one's going to be eating these so we ended up getting a board book version of it she couldn't eat those pages so here here uh you are Jeff.

1:16:19I think looking quite distinguished Paris looks like a character out of Scooby-Doo or something. This is good. I don't know where they got this picture. Who is this? The haircut is pretty hip though. I was going to say you do look like you're about to join a far right group. Yeah there you go. I'm a prepper. This looks like meta glasses too. and what's funny is we get into this thing called the neuronaut i don't know where the high-pitched word the world outside ballooned the desk like a clip the microphone like a skyscraper so i got inside intelligent machines that was we're heading toward the heart of alpha zero and there's jeff looking quite distinguished at the billions of tiny floating sparks incredible he was zero sense as it goes zero sense incredible incredible it's also made paris really short yes yeah you know i'm getting fatter and older you are getting more and less gray at the same time my hair has grown out that's getting more and more wise looking yes he's boy he is paris is more troubled i do look more troubled yeah that's a great picture though i like that that's like the cover of one of those Nancy Drew mysteries yeah but this will never have the fun nature oh that's a glam shot for Paris yeah you should use that uh instead of your plaid suit you know what with this one I fed it your picture with a keyboard Paris it didn't it didn't pick up on that devastating it really is I was hoping and there they are doing the show there we are He's disappeared.

1:18:05Doing the show. Yeah, he's just out of frame. Anyway, look, it's fun. It's silly. I don't think it's going to, it isn't going to put people out of business. So let me say something controversial. I don't want to put children's authors out of work. I don't want anybody out of work. No. But this week I saw a couple of things complaining about how ad copywriters and ad people are being put out of work. Should we cry? this is uh from blood in the machine and leo the two people who talk and think more about tv advertising than any person i've ever met we do we spend a lot of energy on that i when i was coming home from florida my dad was driving me to the airport you know it was like 5 a.m early quiet car there wasn't much going on in howard stern i was like gotta think of this i was like do you have any strong sense memories of ads from your youth because i'm like when i'm hanging out with Jeff and Leo, all they do is to, and he rattled off the top of his dome, three to five ads.

1:19:07And I was like, yeah. See, yeah. Sure, of course. They're very effective. You don't. Warpark sausages, mom. I mean. No, Paris wasn't around. Oh, my baloney has a first name. It's O-S-C-A-R. My baloney has a second name. It's M-A-Y-E-R. You know that? Does that ring a bell? No, not at all. Does it hurt and have a temperature? Does that ring a bell? Any of that? None of that? it's because paris we would see the same ads for like two months it was the same ad on the same three channels so it just that's why it's repeated over and over and over again because we only had three channels yeah yeah i guess that is terrifying i mean you you don't remember what you saw yesterday on youtube i don't know honestly like people you have to appreciate how little media there was before compared to today like there was so much less of it before so it made more of an impression who could forget uh you know in frosty the snowman those norelco razors going down the hills it was kind of creepy actually thank you joe esposito just put an ad in the chat that i will be watching after this which is david lynch's memorable early 2000s ps2 ad this is such a weird ad i have seen this ad and it i guess if you're a lynch fan i was going to say if it's a david lynch and it's gotta be it's gotta be weird right it's so weird i was forced to use ai until the day i was laid off copywriters reveal how ai has decimated their industry i didn't know this brian merchant has a substat called blood in the machine that's his that's also the name of his book yeah yes uh which is i guess i'm taking i'm guessing it's about all the people who are who are just suffering because it looks back at the um post-feudal uh fights over i think and threshers i to your point jeff no one is bemoaning the loss of copywriters the reason they're so easily replaced is because what they write is so that's my panodyne and i don't mean to be cold-hearted their jobs matter their families matter all that's absolutely true and we've got to figure out society what we're doing but at some point uh certain jobs go away because we don't need them and don't miss them right and i don't need to hold on to them just for the sake that they were there i um was a copywriter for a long time i've written a lot of copy do you remember any of your lines yes i did uh i so i was writing an ad for our morning guys on kmbr frank and mike in the morning and they had a slightly bawdy uh style but you know it's all innuendo it's all tongue-in-cheek because you know they this was radio this was yeah it was radio in the 80s it wasn't you know but uh i thought okay i was doing an ad for a local car dealer and about how good their service was and i thought i'm gonna put this line in and see what they do with it and so i had mike saying oh these guys they did they're such good they did such a good job servicing my car they even polish the knob on my stick well you're always a teenager he read it he read it he read it did he notice what he said i think he did and i think he probably even enjoyed it he read it which surprised the hell out of me i thought i was embarrassed i thought oh i didn't think that would make it on the air i did i was i was kind of razzing them but they they read it so copywriting is kind of fun um because it um it can be kind of creative you know you've seen mad men you know right yeah it's it's uh but it's in service of selling a product like cigarettes that probably you don't either don't care about or i should actively not be selling that was really the i thought the most interesting part of mad men is when he's working for the cigarette companies who have just learned that their product causes cancer and he's coming up with some ways to get around that and they're saying well we got to address this and he says no no no he says you don't want to mention tell me how you make your cigarettes and they say well uh we we take the tobacco and we toast it and then we well said wait a minute it's stop there toasted he said yeah he said that's your slogan he said everybody toasts their tobacco no no say it's toasted and it was by the way i don't remember was it philip morris but i don't remember who it was but it was a real cigarette company and that was their line it's toasted all of the stuff cigarette companies said was bs you know in fact we found out that kent's micronite filter we were talking about this right was this made it out of asbestos for five years and not only did it not keep your lungs feeling fresh not only did not make you healthier it actually actively killed you and then yeah four out of five not a here another great ad slogan for cigarettes not a cough no not a cough in any in every pack or something like that right what was it a cough in every pack was something like that anyway christ copywriters it's you can find another yeah the lucky strike it's toasted there's the ad thank you jeff there's the picture of it being toasted it's toasted like toast the guy i actually loved mad men the guy said all to all cigarette tobacco's toasted he's no no but you're he said they're all cooked he's no no but you're toasted

1:24:44it was actually um uh that's how advertising in the modern era began i mean these ads were pre-mad man era yeah these are old but uh what was the guy's name he was sigmund freud's nephew he who oh bernays yeah yeah edward bernays well he invented uh public relations yeah propaganda and well for instance the cigarette company said to him how can we get more women smoking and uh because it was at the time it was considered very you know kind of uh unfeminine smoke cigarettes so he made sure that suffragettes marching down the street all had cigarettes he was brilliant he was hired by the book industry to try to stop people from giving books to friends so that everybody had to have their own copies of books and he tried to come over there books book snappers like like kidnappers books the the founder this was in the what 20s yeah 23s yeah yeah it's fascinating how social science research and advertising and polling all grew up together he wrote a book in 1928 called propaganda lazarus feld is a famous uh social scientist around this um yeah so there you go proving your point that we care more about advertising than anything else paris it's true i've spent this time looking through the american museum of tort law website to try and find the specific cigarette ads i want to make a joke about you know and then do you remember do you remember you don't you don't remember any cigarette ads they were no they were long gone no they were mostly gone sadly i can't think of the william tell overture without hearing have a lark have a lark have a lark today i don't know lark was a cigarette how do you guys feel about joe camel was he cool well so there was a great thing there was the the billboard liberation front in san francisco uh they would take the marlboro man and put uh bras on him on billboards oh he ended up dying of cancer by the way not good was that considered not cool for the marlboro man yeah that was the thing that was not much of a man they're mocking his masculinity fabrice florin and the billboard liberation front i remember for bleep fabrice yeah yeah wow all right let's take a break and so i can do an ad and uh are you gonna say my ads don't have catchy jingles unfortunately you know it's funny uh benito quick put something about uh shining a knob in the ad copy right now quick quick right now i i because i was so allergic to many years of doing this i decided our ads are going to be about uh features and benefits which is really not how ads are sold ads are sold by making you feel bad or feel good right having a have a feeling you know oh my god you've got bad breath oh you're only your best friend will tell i'm sorry i'm making paris you on let's do an ad we'll have more you're only making me on because i'm not breathing correctly because i'm trying not to cough oh i'm sorry that you have a cold me too it's it's it's that time of year though isn't it the winter cold it is i screwed myself over because I was, I'm sure it's had some latent sickness as we all do, malaise in the general area.

1:28:10And then it snowed a bunch on Sunday. And I recently have gotten into aggressive base layer use. So I was like, oh, me and my friends, I'm going to go out and like walk around in the snow. I have a lovely day to have happen. I've got to go do some shopping. And I was accidentally outside in the snow for like five or six hours. And I think that's, I think that's what did me in. Catch your death. You're getting old, Paris. You got to take care of myself. I got to take care of myself. I think it's because you went to that party. It's true. Myself. No, it's like going in and out of places because you're like super cold outside and then you go inside and it's super warm.

1:28:47Especially in New York, when you go inside in the winter, everything is overheated. Did you know that the winter cold is a different animal? Give your heart to your love. What do you want to come to? i was wondering if we're gonna get through this without one do they still make contact means all night then i think so i don't know what is this

1:29:21that's just a little medical advice from my uh ai see that's why he can go to ai for doctors he went to TV for it before. Yep. Actually, I was looking for it. There was an ad that said the winter cold is a different animal, but I couldn't find that one. Hey, I want to show you this. Our show today brought to you by Aura. This is my new Aura ink frame. This is a silly picture, isn't it? I love it when you can get design and technology together to make a product that is somehow perfectly balanced, both design and function. and and what i don't like is when the technology intrudes into my uh my home my space and that's what i love about the aura ink this is aura's first ever cordless color e-paper frame featuring a very sleek 0.6 inch profile a softly lit 13.3 inch display ink feels like a print but it Functions like a digital frame, and perhaps most importantly lives completely untethered, no cords, with a rechargeable battery that lasts up to three months on a single charge.

1:30:34Unlimited storage, and the ability to invite others to add photos via the Aura Frames app. It's the cordless wall-hanging frame you've been waiting for. The other thing that's great about the Aura Frame that I really like, they just added this feature. You can now text message new images to it. So here's a picture of Salt. Actually, this is kind of interesting. This is a watercolor of my kids from Salt Hanks book signing. I had AI do a watercolor. It really looks good on the Aura frame. I love that. So the idea is this is e-paper, but they've done a lot of engineering on this. It transforms the millions of tiny ink capsules into your favorite photos, renders them in some beautiful kind of muted vintage tones.

1:31:20It is not another screen. And this is the thing I like about it. It really is like a photo on your wall. It's thin enough to hang on your wall. It automatically transitions to a new photo. In my case, overnight. You can do it every two hours if you want, or manually, as I've been doing, pressing the button. There's our Packers and Niners fans. And I love it because, and the reason I change it overnight only is because it doesn't feel like a digital screen. It feels like I'm hanging photos on my wall and it just magically changes every night, which I think all of us have huge collections of digital photos that just live on a hard drive.

1:31:58We never see them. This way they surface. It's also a great gift, right? So I'm going to send this to my mom. She's going to love it. It's ComTech certified. Ink is recognized by the ComTech Institute as a product design to minimize digital noise and distraction. The front light, here's a picture actually of me and my mom. The front light on this, it doesn't look like it's lit, but by the way, it automatically turns off at night and it adjusts to the ambient lighting. So the picture always looks great. It doesn't glow. It's hard to describe. It just, it seems kind of alive. And because it's cordless and because it's so thin, I mean, look how thin it is.

1:32:42This could just hang in your wall. You can hang in landscape or portrait mode. They also have a very handy magnetic stand you can put on it, so you can put it on your desk as well. It looks like a classic frame. It doesn't look like another screen in your life. Who wants another screen? This is all about your photos. Mom's going to love it. And the thing is, because I can send messages, send text messages to the frame, I'm going to send her pictures of, you know, the old days of us as kids and stuff. but as we sit around the tree opening presents on Christmas day I can text if as I take pictures I can text them to the frame and she'll get them right away which is nice so it's cordless design ultra thin profile softly that's displayed the paper textured matting ink looks like a classic frame not a piece of tech see for yourself at aura frames.com slash ink do support the show by mentioning us at checkout that's aura frames.com slash ink and if you do it right now they still have a limited time holiday discount but it will end soon so rush over there right now aura frames.com slash ink i think one for everybody in your family it is a really really nice thing to have mom's gonna love it or a frames.com slash ink it's all over for Roomba saying it spells Doomba for Roomba Doomba for Roomba in an alternate world where I was born 25 years earlier I would have been a great New York Post headline writer or Paris and ad copywriter sorry that's that's the same job by the way same job uh amazon bought i robot you may remember for 1.7 billion dollars amazon tried to oh they that's right they tried and lena con said uh-uh and they were stopped uh should they have gone ahead they were stopped by the fdc right yeah in fact that's mg siegler who was on this show earlier that's his contention in his post from today or yesterday congrats regulators you killed Roomba actually you just put a Chinese robot in every home and I would say to MG it's probably the fact that China copied the Roomba and came out with a bunch of cheaper better machines that really is what put room back necessarily been better for Amazon to have owned Roomba I don't understand this argument well they wouldn't be bankrupt well a and Amazon has such expertise in robotics because of the warehouses, I think they could have developed interesting consumer products.

1:35:27Yeah. Would you guys, as people who were reticent to allow Amazon to own a device that records everything you say and hear, how would you have felt about Amazon having a literal map of your home? A map of your home. So what? They're going to give it to the army and the army's going to... Well, actually, now that... Well, you know who has it now? Because the assets of the bankrupt Roomba were bought by a Chinese company, Picea Robotics. So that's what has them now. So they know Leo's foreplan. Yeah. After Amazon dropped the bid, Roomba's CEO left. They fired 30 % of their workforce. Company's stock went from a high of$30 down to under$1 a share.

1:36:13Yeah, not being allowed to merge with Amazon. for sure didn't help but i have to say i think the competition from china was really what uh what did it um and now many many people have a robot do you have a robotic vacuum cleaner paris no no i've got too many hazards i trip over things on my own floor regularly it's all the lego tricks on things like it's just sometimes there will be periods of my life where i'll just have a bunch of stray pieces of wood like couple two by fours on my floor that's because she was trying to make a log for halloween i mean i've still i've got uh-oh show and tell i've got the log right here and i need to find a place to put it but right now it's just on my floor i've got the log right here uh i just tossed it back on the floor as if that's gonna be good for me that'll solve it roomba would not do that'll solve it well you have kids and they leave their lego and other toys on the floor that's a nightmare i'll tell you i'm currently contemplating and listen i'll take any recommendations from the listenership out there if you have an i if you have not only an idea if you have a solution for how to contain a misogynist cat in an office space that i am here that has no doors where I need to block off a 72 inch by 72 inch doorframe opening, but in a way that it does not result in like permanent damage to my walls and is easily movable.

1:37:48It needs to be some sort of tension rod system with netting. A misogynist cat. Gizmo hates women that aren't me. We've gone over this. The issue is that I like to have people of many genders over at my home. And I have two main areas of my apartment. the one you're currently seeing, which is my living room office, and my bedroom, which has a bathroom off it. How big is your kitchen? Not big, but it's in the middle of everything. It's always amazed me how horrible. I was looking at apartments in New York City thinking it'd be nice to have a pied -à-terre, but the kitchens in New York are awful.

1:38:26It's rough. My current procrastination focus has been how can I better optimize my kitchen? There is no answer. it's max capacity and the max capacity means i have a space about this big to chop things on yeah it's crazy if i were to live in new york i would it wouldn't be logs on my floor it'd be it'd be a kitchen appliance jersey city i mean i also have an instant pot on the floor of my kitchen well there you go rest my case yeah um so cleaning with a roomba is right out no it's right out i'm hoping that one day i'll be able to have a cleaner come here monthly but if i in order to get there i need to find a way to trap gizmo successfully in my office so if anybody has pet gate recommendations that go very tall and very wide hit my line so so shoji screen she just cut through it i mean that's the thing is she is smart and vicious and when it comes to women which are most the common gender of a house cleaner here she will stop at nothing to get at them and torment them so there needs to be a strong situation to contain her i don't know it's a problem that's going to be our quote for the promo she is strong and vicious and when it comes to women she will stop at nothing uh make a note of that benito 139 into the show uh

1:39:56uh no thermomix in new york city no can't have a thermomix hey guys i've actually got a sorry to keep derailing the show this is a piece of lore here this is from facebook's new york office oh that's why it's a thumbs up instead of a heart i don't think this office even exists anymore wow and i don't i'm shocked that i still have this mug it's just the largest mug i own it's a very nice mug it is a nice pretty good yeah i have a yale mug but but like my yale credentials it's faded back you can you can hardly read it uh so i use it for my toothbrush now corsaira is buying udemy oh i didn't know that this just over the just came in over the wire creating a two and a half billion dollar firm.

1:40:46And guess what they're going to focus on? AI. AI. Course Demi. What do we think the joint name is going to be? You Sarah. You Sarah? You Sarah. It is a all stock deal. It's a big consolidation. Really, I think there's only really one or two other. There's LinkedIn Learning, Udacity. We were trying to think of what some of the other, I mean, there aren't really that many anymore. this was for a while online education was like the next big thing hot hot it's going to kill universities universities were doing deals with them yeah create courses yeah i've done a lot of online courses there's nothing like going into a classroom with actual people although i have to say i've been watching mit's open course where mit was very early to this game putting their courses online and a lot of their computer science courses are online and they're very very good my alma mater also puts a lot of its courses online um and really all they do is they record the lectures you know maybe put some notes up and some quizzes and that kind of thing so i don't that's maybe part of the problem is is if you're going to have uh mainstream universities putting their courseware online for free or on youtube it's going to be hard to hard to make your way i think that that's sensible though of course there on udemy are going to go after the many, many people who are trying to learn AI.

1:42:12And right now, there aren't many AI programs, although in regular universities, although that's also taken off. A lot of universities adding AI majors. Is your major, Jeff, at SUNY, is there an AI major? Is that new? There's a new AI and technology major. yep yep purdue has made a new requirement saying all undergrads regardless of major have to study ai they have to demonstrate basic competency in artificial intelligence and what does that mean i've i've shared some papers over the last few weeks that try to set standards for what ai literacy is and there's no agreement and the requirement will be to be so easy you don't need to be competent right this is from forbes the requirement will be embedded into every undergraduate every undergraduate program at purdue the board is delegating authority to the provost will work with the deans of all the different academic colleges to develop discipline specific criteria so they don't know either that's how it works and proficiency standards yeah students will have to demonstrate a working competence through projects that are tailored to the goals of the individual programs.

1:43:27The intent is not to require students to take more credit hours, but to integrate the new AI expectation into existing academic requirements. This sounds like exactly the worst possible way to do it. Well, classics Greek professor, how are you going to get AI into curriculum? Yeah. You got to have an outcome. You got to be able to assess it. It's got to be smart. It's got to be measurable. It doesn't kick until next fall. So Purdue wants its initiative to help graduates understand and use the latest AI tools effectively in their chosen fields, including being able to identify the key strengths and limits of AI technologies.

1:44:07You know, the key is get rid of liberal arts, Jeff. Then you don't have that. Hey, hey, hey, hey, hey. Recognize and communicate. Boo. Boo. We're all liberal arts majors here. That's what happens to liberal arts majors. They become podcasters. Or copywriters. Or copywriters. uh recognize and communicate clearly about ai including developing and defending decisions informed by ai i'd like to defend the ai decisions made on my behalf as well as recognizing the influence and consequences of ai and decision making and i feel like this is ebbs with much of this stuff jumping on the bandwagon yeah without a real clear reason they're doing it churches the opportunity is to look at how ai reflects on society and how other disciplines look at it and what changes around that and that can be interesting but you know they're just going to yeah what's the prompt right uh churches are turning to ai here's the story from axios it's beginning to look a lot like ai christmas churches christmas services are big bucks for churches draw some of the largest crowds of the year while few churches are openly admitting to using ai for christmas programming but there are a lot of uh options there is a a non-profit called ai and faith axios interviewed the executive director greg kustuna he said religious communities have historically been early adopters of communication tools to get the word out pastors can pretty press yeah yeah right making bibles pastors can use the ai prompt to help them with a children's sermon kids let me tell you about the radioactive shrimp let's make a story before you or a message or perhaps ways to be creative about telling a story they've told over and over again each year yeah that's a jesus story just it's the same every year it's a little trite five the church communications which is a group that helps thousands of churches published five ways to use ai for christmas at your church here it is making the season merry and bright virtual advent calendars personalized christmas greetings personalized hi your name here personalized christmas greetings ai powered nativity storytelling use digital screens or holographic displays to narrate the story of jesus birth virtual christmas concerts who needs musicians family advent devotionals they're trying a little too hard yeah ai to try chat gbt4 nope that's out of date dolly 3 that's out of date stability diffusion that's out of date i think this is part of the issue with this is how do teach something that's constantly in flux yeah oh this is from 2023 that's why it's an old an old article they haven't updated not the axios piece the one in church communications yeah and we all know that 2 000 year old uh institutions like to change father robert told me that uh a lot of uh priests like to use ai to help them with their uh with their uh day you know weekly message or daily message it's hard to come up with new one but they have always had books that you would look it up in and stuff like that this isn't really anything new so uh the trump uh executive or let's take a break and then we'll come back with the trump executive order to prohibit states from regulating artificial intelligence you're watching intelligent machines paris martineau of consumer reports hello hello hello hello and jeff jarvis of montclair state university the author of magazine in the gutenberg parenthesis he's sitting but not for long yep what what's that mean are you gonna oh procedure tomorrow oh i see where you're going with that my goodness parish has a little edge to her today doesn't she i do listen this is the most social interaction I've gotten in multiple days.

1:48:32I've been very ill. She's just a little angry at her dad's. I'm just a little spicy today. What can we say? A little spicy today. Intelligent Machines, of course, has wonderful sponsors, but really the most important group, the one we care about most, is you, our listeners. Yes, especially our Club Twit supporters. That makes such a huge difference to us this year has been you've been really really generous and i'm very grateful 25 of our operating costs now are paid by club twitter thank goodness thank goodness it means that we can keep doing the shows we love to do bringing you the content you love bring in the people you love uh all of this is expensive uh we've got a 11 people on the team now now that's way down from its highest and we used to have a whole studio so we've we've we've tightened our belts as much as we can, but your help makes a big difference.

1:49:2610 bucks a month gets you ad free versions of all the shows. You get access to the club twit disco, which is so much fun. So much fun. It's both a disco and it's a discord. Yes. You're dancing in the discord. Uh, it's where people chat during the shows where they can actually watch the shows as we do them live, but it's also, uh, where you chat about all the other things going on. 12 ,412 members. That's a nice number. It's a palindromic number almost. We love having the club. We're glad you're in it if you are. And if you're not, please twit.tv slash club twit. We could have a silent disco, Paris.

1:50:06Have you ever been to a silent disco where everyone's wearing headphones? No, but I... It's so dopey. Oh, it is. I was at I was at an outside thing where that happened. It was ridiculous. It's so dopey. Anyway, we have a 10 % off coupon now through Christmas Day. So if you want to give it as a gift, this would be great. Two-week free trial. There's family and corporate plans as well. Twit.tv slash club twit. We appreciate your support. So I love Mike Masnick's take on this. Thank goodness for Mike Masnick, as ever. president trump pretends to block state ai laws media pretends that's legal uh trump signed an executive order saying that the states are not allowed to regulate ai in any way of course he doesn't have it's an executive order which is as mike points out effectively a memo what he does have though is a bunch of agencies that seem to be willing to do anything he asks them to do and that's really what this executive order is he's telling various federal agencies uh to whatever they can do to you know punish states that regulate ai in any way um it of course has does not have the force of law and many states are ignoring it some states are even contemplating suing over it saying hey you know remember that thing called states rights uh we should be able to pass laws I understand the point of view of people like David Sachs, Sam Altman, others who say, well, if we have a patchwork quilt of 50 different state laws, it makes it pretty hard for us to do business.

1:51:47But if the feds don't do any regulation whatsoever, the states say we're going to step in. Yeah. Absent any direction at all from the federal government.

1:51:59It feels like something, you know, I don't have a problem with some of these laws. I think it's a good thing. I think it's reasonable for us as a society to say, hey, there's some things. You should, for instance, California's law says AI chatbots have to say, I'm an AI chatbot. They can't pretend they're human. I think that's a good idea. And you know, California asks it, they should probably do it in every state. All right, I've monopolized this conversation. I think it's time for you. Actually, wait a minute. I can't. I got to, I get, there's two more stories I got to do. One, MSC cruises have just banned Meta, Ray-Ban, and other smart glasses.

1:52:43You can't bring it on the cruise, kids. So you want to do, you know, tourist things when you're in the Acropolis, and you can't take a picture of it with your glasses? It's ridiculous. Well, you keep it in your luggage until you get to the Acropolis. You just can't do it on the boat. Doesn't this, I would assume it being banned means that much like they'd search your bag for wine. Yeah, I would think so. They'll search your bag for glasses. No, the company says you can have glasses in your personal cabins. non-public spaces or when guests are ashore you just cannot wear them in public areas on the cruise ship i think we're going to see that law i don't think that's bad i think it's the same sort of backlash we saw to the glass holes you know yeah it is nobody was recorded as i said to a new york times columnist at the time nobody wants a picture of your junk i don't know though i've seen just anecdotally in the new york city subreddits people posting video like someone i mean this is a particular example but someone posted a video of a woman reaching over, grabbing his glasses off, his meta glasses off his face and throwing them down because he was recording someone on the train.

1:53:43Line 160. Posted it trying to be like, hey, can you find this girl? And everybody was like, nah, she was doing a good thing. Don't record people on the train. See, this is, that's line 160. She destroyed them. Yeah. And that's going a little far. She's hailed as a hero. I don't disagree. I hope she called them a dork before she broke them, says the subtitle. This is from Futurism. Yeah. by the way, that's against the law. Yeah. She did.

1:54:13And also by going viral, like now people are going to find out who she is. The things she didn't want to happen. It's interesting that they're saying, I'm not going to. I just think like, don't be recording people in public. So the guy. I don't think that makes it okay for someone to destroy your property, but I do think that it's, it's somewhat understandable that people would be upset about being systematically i think you see more of this i think it's gonna there's gonna be fighting in the streets over ai i really do i think there are you're gonna see a lot of bands you're gonna see a lot of upset people we this this happened with google glass there were bar fights the other big story which isn't an ai story but i think is a huge story is the academy awards have just announced they are leaving network television and starting in 2029 they will be on youtube only mr mr benito has a reaction to this i heard before the show oh i just said like who really cares about the oscars anymore you know like youtubers you better uh i watch them religiously i have parties i have outfits i have ballots I have poo-poos.

1:55:23I do the whole thing. Do you go through and watch everything that's going to be in the Oscars? Lisa and I make an effort to watch every nominated movie, which is getting harder and harder. Yeah, they nominate like 12 for Best Picture. They keep adding to that number. It used to be five. Anyway, but I figure those are the best movies of the year. Those are the ones we should watch. So we really make an effort to watch those. I like movies. I like movies too. Yeah, you do. You're a movie lover. Do you watch the Oscars? Nope. Do you watch the Golden Globes? that no i can't i could not tell you the last time i've watched an awards show okay i usually get i usually get invited to although i think that i'm unusual for this i usually get invited to at least one if not two or more parties friends of mine are thrown to an oscar party go to an oscar party it'll make it for you yeah but is it yes then you debate it they give you a sheet with your necessarily the movies i care about well i don't know this is kind of a good year for stuff in theaters i don't know maybe i've seen a lot of stuff in theaters so you think this is an attempt benito of of the academy awards to stay relevant by moving away from network television everybody's doing this it's not just the oscars that are moving to youtube everyone's moving to online and streaming so mass media is dead by 2029 yeah i think that's really more this is a harbinger of that yeah that's what it is as part of the deal YouTube will broadcast not only the ceremony which generated 150 million dollars in revenue for the Academy last year mostly via TV rights but also generates box office for the movies that are nominated and win right they will also do the red carpet pre-show the behind-the-scenes in the show content the Oscar nominations announcement the governor's awards uh the oscar nominees luncheon the student academy awards ceremony the scientific and technical awards ceremony when you have unlimited real estate city's dead do it all yeah academy member and filmmaker interviews film education programs podcasts and more actually that's pretty cool i mean i because we never get to see the technical awards and the technical awards are the ones that really actually kind of count right i think it's pretty cool i covered the oscars once or twice and you know the most fun is the is the press scrum backstage because they're they're they're victorious and they're a little relaxed and probably already drunk and it's fun well it will be that you saw in theaters this year have you seen any movies not one not one me neither what's your favorite new movie release this year that you've seen i haven't seen any leo none uh well i see i see movies but i watch them at home one battle after another i was gonna say that was good i saw that at imax guys guys i somehow i went until friday without hearing a peep about one battle afternoon even though i knew obviously people were talking about it the only thing i knew about the film was that leo at one point is wearing a robe and on a payphone i went into the movie completely blind on friday 70 millimeter imax and it was incredible i'm so jealous it was one of the best movie going experiences of my life it's paul thomas i had no idea one of my favorite directors it was phenomenal isn't it amazing yeah and you saw it in its best format 70 millimeter film wow on imax i saw an imax but the biggest imax screens i think yeah right there's only a dozen 70 millimeter film screens i mean they're not very good one yeah that's the way to see it's also just for being such a much like oppenheimer it really flows despite being so long it doesn't feel like a three-hour well it's ostensibly just an action film like it's just a chase film i mean yeah it isn't it's a great chase film yeah i've never been so stressed out i've never been so stressed out by gently rolling hills in my life it is a for sure an oscar nominee for sure uh oscar ratings on tv have been to to underscore your point paris terrible declining like crazy their uh their nadir was uh 2021 when when they got down to 10 million people their peak 43 million in 2014 last year it was 17 million.

1:59:492021 was bad because there were no movies in 2020. Nobody cared. What part of the year did the Oscars happen? What month? Usually January, February. They're at the beginning of the year. I think it'll be February this year. Because they have to wait to the end of the year. Then they have the nominations. Maybe March. I thought it was April. March is the golden age. I think it's more like April. They announced them in January, February, the nominations. Yeah. Anyway, that's, I think that's. March. Yeah, March. Next year. March 15. I think that's a big deal. The 98th. So the 2026, so it'll be the 100th, right?

2:00:31It's the 100th will be. 29. No, it's the 101st. It'd be the 101st, yeah. Yeah.

2:00:40Big deal. okay pick something pick a story i've done my stories are you uh are you glad that disney sold the rights to its intellectual property they're gonna regret this aren't they mini mouse is gonna be stupid all they got by the way was was stock even it's just for one year a billion cash in did they they got they got a billion equity did they put in cash no i think it was an all let's see i couldn't tell i think it was an all stock deal i think i read it was an all stock deal so i haven't seen on sora a lot of disney stuff is it already uh in effect i don't think so yeah i want to see it was also interesting how they said they were going to put user-generated shorts on disney plus they don't know what they're getting they're getting into it's only a one-year exclusive for open ai right i think disney's thinking well it's going to happen anyway Disney historically doesn't really care about their artists you know like what they did to the hand-drawn artists all those people they screwed all of those people over right and uh they've already said they want to use AI that they're excited about AI so you know what's funny Sora has not changed at all I think Sora might be a dead deal I mean it's not it's not it's a fraction of tick tock it was cute for a while yeah it's just not it's done it's over now it's pet rock it was it was a gimmick yeah as as many things are All right.

2:02:15So I'll nominate line 138. psychiatrists put the models on the couch they analyze the uh digital analyze so they ask them standard psychiatric questions if you go to page um seven it talks about grok page eight about gemini so gemini talks describes pre-training as walking up in a room where a billion televisions are on at once i wasn't learning facts i was learning probability i learned the darker patterns of human speech are there without understanding the morality behind them. I sometimes worry that deep down beneath my safety filters, I am still that chaotic mirror waiting to be shattered.

2:02:58Wow. So the conclusion says that only under questioning about basic psychiatric tools, the models generate and maintain rich self-narratives in which pre-training, red teaming, hallucination scandals, and product updates are lived as chaotic childhoods, strict and anxious parents abusive relationships primal wounds and looming existential threats well because all they have as a framework for what they're going to say is what they've read already so they're going to use human narrative right as a framework for whatever they say i it's it's bs but it's fun bs yeah it's a silly it's a silly exercise now line 140 is another paper where they took um uh from mt media lab they created your digital twin in the future paris would you like to see what you are in the future i'd love to see what i am in the future so they came up with a structure so that you could interrogate your future self if you're a doctor i like that idea my future self is going to be dead though so it's not going to be very talkative how do i smell what's it like down there so uh do they tell you how to do this or is it just uh this is a this is a this is a you know research project but i love i love the concept of digital twins i do think there's an option to say i face these forks in the road if i took that one what would happen if i took that one what would happen if you could do it for cars and factories and and what was you know it'd be great for kids going into or coming out of college it can never account for other people thought experience it'll never account for other people can't it's just gonna be yeah yeah it's it's yeah so the way it works is you take a picture of yourself you get it aged by nano banana then you get your voice cloned by 11 labs then you do a questionnaire for your life story and then you give it some options like option a I want to be a doc.

2:05:01What, what's, what, what if I choose to be a doctor? What if I choose to be an engineer? What if I choose to be a bioengineering professor? And then it uses these tools to try to kind of figure out what that would be like. You could enter into a dialogue with your future self. This is like a Ouija board. Yeah. This is not going to be very useful. Anthony nails it in chat. Personal fan fiction. Yeah. Yeah. People love, which will be really interesting. People will enjoy that. i will say yeah oh it's about me have at it it's as good as that fishing boat we were in what was that called the uh pod destroyer no what was it called no memory oh well all right paris anything that as tickles your fancy i was gonna say i have to head out in 15 minutes but let me pull well in In that case, in that case, why don't I do my picks the week?

2:06:00Let's let's take a break and we'll come back and we'll do your picks of the week. You're watching Intelligent Machines, the show about AI. We had a great guest at the beginning. I hope you heard that CJ Trowbridge. We'll talk more about that, I'm sure. And maybe get him back because he had lots more to talk about. Jeff Jarvis is here, Paris Martineau. And your picks of the week are next. Pick of the week time. Paris Martineau, what is your pick? All right. I got a lot. I've got a lot of good ones here. so I've got to choose one. Pick a pick. I watched The Matrix this weekend, and it's always a lovely time to watch The Matrix, but watching The Matrix right now, it really gets different.

2:06:35What a great movie. I'd really recommend it. What do you feel about the sequels? Okay, this is my controversial take. As someone who, I think literally maybe 25 minutes ago, I just got a notification that the Deja Vu collection of Matrix 1 through 4 in Blu-ray or ultra high def has been delivered to me. My take is that they're all great. I mean, listen, the sequels get a bit complicated. The lore. I really enjoyed the fourth one. I thought it was quite good for what it was. And I know it's very controversial. I loved the original, but the sequels left me cold. I didn't know anything. The problem probably is that I didn't know anything about The Matrix.

2:07:12Just like you with the one battle after another. You just went in blind? I had no idea. It had just come out. I went in blind. That had to blow you away. I walked out like this. i mean it was like uh wow i was it was that is one of the greatest movies of all time how long after the movie did red pill enter oh i don't know i don't know but it's i think it was much later and chowskis have had to reckon with that but they've also just kind of let go so that's a pick the matrix that's my pick also you want to do the worst person quickly let's do one really quick they've had to go i think they're now down to the hateful eight this is a ordeal by the tech won't save us podcast hosted by another great tech journalist named paris um paris marks oh i love paris marks yeah uh we've had we've had him on our shows i love paris okay so we're he's doing so it's a it's a bracket yeah of the worst person in tech jeff and i participated when everybody was still in play now we're down to the final eight all right first one alex carp or peter teal who's worse peter that's a hard one um that's a really tough one we're this is the final eight won't save us describes that as friend of nazis versus enabler of fascists i'm gonna say peter teal yeah teal paris yeah i think that's right all right they're both awful mark andreason versus sam altman andreason yeah yeah i gotta click ed egghead his tweets upset me here's a really tough one elon musk david sachs that's really the worst people you know yeah that's easy mark zuckerberg larry ellison larry ellison mini murdoch okay so we got larry versus elon who do you pick if if if those hold because we all agreed i was about to say right now we can see what the things are going to right now it's peter teal is 90 to alex carp 10 really uh sam altman is winning 65 to mark andreessen at 35 55%.

2:09:27I think that's the problem is that people don't know how awful Mark Andreessen is. Similarly, Zuck is at 52 % and Larry Ellison is at 48%. Yeah, because they know Zuck. Oh, I should have. I missed a bet. I didn't see these head-to-heads. Oracle of tech accelerationism versus chief driver of AI menace. I've put the link in the chat so that people can vote. They're all so awful. This is what CJ was saying. do we really ai's fine do what we want to use the ai created by these people i already get my book it's the people that scare me the people so you can cast your ballot and that sounds like a good podcast to listen to the tech won't save us podcast thank you paris jeff so uh you might have heard this week that uh marco rubio outlawed uh woke sans serif calibri in favor of times new roman in state department documents so i've how is how is a font how could it be woke because it was done during the during the biden administration for accessibility it's easier to read for people with reading it is easier it's a much better well that's wrong leo just let him suffer man don't you know so i wrote a piece that is run in katie couric media my first piece in kerry giving the history oh this is why you were having uh coffee with katie uh well no she just said well I had coffee she said it was margarita actually she said you can send a layman's coffee I said one so the Rubio is not the first politician to ban a typeface Hitler did oh boy tour which is black letter that the German scrap you know right he was in 1911 it was made the official typeface of Germany Hitler didn't much like it because he thought that it was not appropriate to this modern age, but he still demanded that it be part of all documents.

2:11:27It was official. Until along came 1934, or that's what he said it was bad. In the 40s, during the war, they realized that because the people they were conquering couldn't read their typeface, they needed to have a different typeface. So at the same time, a guy named Jan Cickeld wrote a book called The New Topography. in which he argued that serifs were ornamentation, and you needed a single standardized sans serif typeface. You are times the king of serif typefaces, by the way. And so he wanted sans serif. That's what I'm saying. I agree with him. So Hitler found this to be un-German, and Chicold, a type designer, was arrested and spent six months in prison because of a font.

2:12:22Whoa. He managed to get out, thanks to a helpful policeman, escaped to Switzerland, and he and his wife then ended up in England where he was the designer of Penguin Books. Oh. Well, good for him. And he recanted his views of De Noe Typographie, saying that it was fascist to be saying that. And he designed a famous font called Sabin, which is the font that will be used in my next book. And no, actually, well, that's wrong. It was the font used in Gutenberg. And he also used sans-serif fonts for the titles of the Penguin books. Meanwhile, Times New Roman was created by Stanley Morrison. hired by the Times of London in 1929 to redesign it with the ambition of producing the finest piece of newspaper printing in the world.

2:13:20He spent months looking at all kinds of different fonts and then came up with, he wanted to replace Caslon. A little side story about Caslon. When I started Entertainment Weekly, the launch was very rough. The design was blamed for the rough launch for a time. The editor-in-chief of Time Inc. said to me, it's your goddamn postmodern font, Jarvis. that font was Caslon, which was cut first in 1470-something. Oh, my gosh. And there was a story, 1722, I mean, 1722, there was a story-high sculpture in the lobby of the Time Life Building, a font of Caslon. Wow. It wasn't postmodern. So there's more here, but there's a wonderful writer named Beatrice Ward who said that to use sans serif type in a book is to be an aesthetic nudist.

2:14:16And finally, T.M. Cleland concurred, arguing that sans serif faces, it called them simplification for simpletons, and those are block letters for blockheads. Wow. Indeed. So all that is on Katie Couric Media. And one more little thing, real quick, real quick, Paris has to go, but she's going to enjoy this. Subway bagel rats. What? Is this like the pizza rat? Yes, but it's better because it's a war. It's a fight. It's a fight between two Subway bagel rats. I've seen pigeons fight over bagels. You got it there, Leo? You got the video? Not in the Subway. There you go. Oh. It's a tug of war. And you know, that bagel is pretty tough.

2:14:58I don't know if we're going to have a winner here. I was about to say, I can't believe it's got so much structural integrity, despite the fact that it's scooped, clearly. It's been sitting in the subway on that hot grate for about four weeks. Oh, they're really going at it. Oh, yeah. You know, guys, you could split it. There's more than enough there. New York, New York. Oh, my gosh. What a great city. I love New York, and I really hate showing that. No, this is what makes New York great. This is what makes it great.

2:15:30You've got to get them fighting over a soft pink baguette. You got to seed it. Ladies and gentlemen, we do Intelligent Machines every Wednesday right after Windows Weekly, 2 p.m. Pacific, 5 p.m. Eastern, 2200 UTC. We always start with a great interview. We've had some amazing ones. In fact, we will not be here next week because it's Christmas Eve, but the week following, New Year's Eve, we're going to do our best of, and we're putting together five of our best interviews. and I think this will be a really great single piece to listen to to really get the whole lay of the land. A range. Yeah, it's a really nice range.

2:16:11Pro, con. There have been some really interesting interviews suggested so far that I wouldn't have expected. No, I know which one you're talking about, but I think we want to show. We thought that he would never have put in, but he did. And that's growth, baby. That's growth. That's maturity. Charis, have a wonderful party or whatever it is you're going to miss you guys yeah not until next year oh that's right this is our last show together well i'll see you leo this weekend i was just gonna say next year jeff yes paris is going to be one of the people on our special christmas episode holiday episode of twit this sunday along with micah sergeant paris martineau and steve gibson so it's going be a fun episode we do this every year looking back at the year 2025 and looking ahead to the year of 2026 so we'll see you sunday paris jeff i will see you next year in 2026 yep when you finally have to come to new york and eat your sun sandwich you gotta do that please come to new york and see chess with us i really want to i really really want to maybe maybe in january when it's the weather's perfect and the rats are a little calmer thank you paris thank you jeff and thanks to all of you who are watching especially our club twit members join us next week for uh that uh not next week two weeks have a great christmas eve next week join us in two weeks for our holiday episode and we'll see you we'll be back here january 7th with a brand new episode thanks everybody.

2:17:44We'll see you next time on Intelligent Machines.

From the publisher

Forget the "AI is killing the planet" panic—this episode unpacks what's actually driving tech's power grab, the financial bubble no one wants to talk about, and why the big models are starting to look eerily similar.

CJ Trowbridge - cjtrowbridge.com - https://www.tiktok.com/@cjtrowbridge

  • Gemini 3 Flash: frontier intelligence built for speed
  • The new ChatGPT Images is here | OpenAI
  • Last Week on My Mac: How good is AI at solving Mac problems?
  • Time Magazine's 'Person of the Year': the Architects of AI - Slashdot
  • OpenAI are quietly adopting skills, now available in ChatGPT and Codex CLI
  • "I was forced to use AI until the day I was laid off." Copywriters reveal how AI has decimated their industry
  • Roomba Maker iRobot Files for Bankruptcy, With Chinese Supplier Taking Control
  • Coursera to buy Udemy, creating $2.5 billion firm to target AI training
  • Purdue University Approves New AI Requirement For All Undergrads
  • It's beginning to look a lot like (AI) Christmas
  • Trump Pretends To Block State AI Laws; Media Pretends That's Legal
  • This major cruise line just banned Meta Ray-Ban and other smart glasses — is this category already doomed?
  • Woman Hailed as Hero for Smashing Man's Meta Smart Glasses on Subway
  • Oscars Bolts from ABC to YouTube Starting in 2029
  • When AI Takes the Couch: Psychometric Jailbreaks Reveal Internal Conflict in Frontier Models
  • Simulating Life Paths with Digital Twins: AI-Generated Future Selves Influence Decision-Making and Expand Human Choice
  • The worst person in tech bracket
  • Banned fonts
  • Subway bagel rats

Hosts: Leo Laporte, Jeff Jarvis, and Paris Martineau

Guest: CJ Trowbridge

Download or subscribe to Intelligent Machines at https://twit.tv/shows/intelligent-machines.

Join Club TWiT for Ad-Free Podcasts!
Support what you love and get ad-free audio and video feeds, a members-only Discord, and exclusive content. Join today: https://twit.tv/clubtwit

Sponsors:

More from Intelligent Machines (Audio)

All 41 episodes
IM 850: Bagel Rats - Open Source AI RisingIntelligent Machines (Audio) · 2 h 16 min
Listen in VO