IM 878: The 1AM Bus to Chinatown - Robots, Bison, and the Art of Savoring Small Stuff

37:49But at least in the case of gratification, to understand that we can get it from almost anywhere. So I was talking today to a Swedish academic named Jonas Iverson, because he wrote a really interesting essay about friction and arguing that AI makes everything so easy. It takes away. I got plenty of friction. I got friction to spare. So is friction necessary to learn? Is it necessary to gain knowledge? Is it necessary to feel satisfaction? We hear a lot of arguments about friction these days. We've heard it lately. I'm kind of against the whole friction discourse. I don't think it's wrong, but I think it's not the point I'm trying to make with gratification.

38:31Because the thing that I just showed you where I was able to touch the type on my book cover, that didn't require difficulty. That wasn't about friction. That wasn't about making my life harder in order to be more involved in it. So I don't think, I think it's understandable to me why friction has emerged as a tempting solution. because technologies have made things so easy. And, you know, your literal smartphone screen is frictionless and you can glide your finger across it and everything seems possible on it. So I understand the motivation for citing that example. But I actually think at least in terms of feeling gratified, you don't have to make it hard.

39:12And in fact, one of the properties of gratification is that unlike happiness, it's easy. it's always available to you. So I have this like textured polo shirt on today and it has this sort of, it's knit and it has this bumpy texture and I can touch it or I can just feel it on my body when I'm not touching it and you can see it and perceive its texture. And it doesn't require us to like, I don't know, like add friction to our wardrobe or something. What would that even mean? It's just, it's just like right here, right, right. Yeah. Like I'm going to wear a hair shirt. It's just right here. And yeah, maybe there are aspects of my life I want to do more with my own hands or I want to be more engaged.

39:54And I think that's where the friction discourse kind of really means to live. But to me, it strikes me as like a way of trying to, again, to divide the world into two categories, ease and difficulty. And everything has been made so easy. Now we need to make things difficult. These kids today. It's like internet life has turned us all into extremists. Everything has to be all or nothing all the time. And I just don't quite buy that. But I also think AI has made my life harder in a good way as much as it's made my life easier. So I was mentioning that I had my countertops redone so that the the the stone guys who put in my new countertop broke the the knob of my range oven off and it's the kind of oven that to replace it i have to replace the whole thermostat assembly which is like tied to the back of the unit and all of this and as i was trying to figure out what to do about it i just asked it i asked chat gbt like hey like what part does my viking range need and you know it helped me find it and then it's giving me some tips about how to go about repairing it if I choose to do so myself.

41:06And like that opportunity, normally I would have just called somebody in or like yelled at the stone guys, I can't believe you broke my oven. But that's interesting. Like now I have the ability at least or the opportunity to engage with that ordinary object in a different way. And I would feel satisfied if I completed it. And I would feel gratified while I was doing it. And I would feel happy to have a working oven again that I could cook food for my family with. And that's not really friction either, right? It's like a different invitation for a complex multifold engagement with the world that isn't just me scrolling either.

41:45It's just like a lot more complicated than we've been to credit for. My wife would say, you're never going to fix this and you know it. I think I can fix it. But even if I can't, even if I can't, the feeling of having failed to fix something, but to try to do so, that's really opened doors for me. I can't remember if we talked about my irrigation backflow last time I was on, but I had a similar problem. Yes, we did. Yeah. So when that happened and the guy, I didn't fix it, but I tried and the guy came out and I had this just like deep, rich conversation with my irrigation repair guy about backflow mechanisms.

42:19And like, I was able to engage with him as a person just much more deeply than I would have been able to otherwise. And I really enjoyed that. I had the same experience with the septic tank guy. Normally a guy I would have stayed as far away from as possible. But instead I actually had a wonderful conversation. He invited me to coffee. Yeah. And you know, one of the things underneath this argument that I'm making is that like, I am fortunate to like be in a socioeconomic situation, which I own my home and have incentives and inspiration to repair. Not everybody gets those opportunities to connect in a sensory way, to experience the gratification of engagement with something that they're invested in.

43:01And that's really too bad. That's one of the forces of dematerialization that I talk about in the book. So I just, I'm ready to post for the Bloomsbury blog now, because you do that to promote your books. Yeah. And my similar experience as a non-academic as a poseur in the palace, when I discovered the joy of research and digging rabbit holes into, or exploring rabbit holes into academic databases and finding citation upon citation upon citation, that's my similar too late in life found joy. Yeah, yeah. I was talking to someone recently about the lost art and craft of like card catalogs and microfiche and microfilm research.

43:47And, you know, that was also, it was pleasurable because knowledge lived somewhere and you got to go to those places and they often had grandeur architecturally. And then you got to be quiet inside them and search for esoteric material. Yeah, but it's better now because you can find all kinds of stuff you couldn't have found. It's definitely better now. It's definitely better now. But also just like one of the motivations for doing it wasn't to do the research, but just to have the sensory encounter with these odd machines. I think the key here, at least what I'm gleaning from it is, yeah, different opportunities for different people, but the key here is to pay attention, to be present with the opportunities you have when things happen, wherever they happen, and your life will be richer and more fulfilling.

44:31And goodness knows, I think we all know that those goals that you spend your entire life trying to achieve often don't satisfy at the end of the day that you get there and you go, Yeah. You know, I don't know, as a parent, you know, like your kids grew up and they leave home and then you miss them. Yep. And the things that you miss aren't like going on a big vacation or writing the college essay or something. It's like hearing the creak of the floor above you. Yeah. Right. It's true. The pitta pat of little feet. Ian, it's such a pleasure, as always, to talk. Ian's website is bogost.com. That's where you'll find the small stuff.

45:14His new book, How to Lead a More Gratifying Life. I think he's given us a great lead into how to do that. You can read his work in The Atlantic. He's working on something right now. He's a professor at Washington University in St. Louis and a game developer. In fact, he has some really interesting books about video games, including Play Anything, The Pleasure of Limits, The Uses of Boredom, and The Secret of Games. James, just one of many books from, one of 11 other books. 11, 11 total books. One of 10 other books. Ian, always a great pleasure to talk to you. It's always great to join you. Thank you so much.

45:51Yeah, we really enjoy it. Come back. You and your Monstera leaves are welcome anytime. Fantastic. All right. Thank you, Dr. Biles. All right. Have a good one. Take care. Cheers. We'll continue with Intelligent Machines right after this. First, a word from our sponsors. You know, Jeff, have I talked to you lately about a new mattress? about a year ago now. Lisa and I looked at our mattress and said, it's been a while, maybe it's time for a new one. Our show today brought to you by our new mattress, Helix Sleep. I guess it's not new anymore, it's a year old. And I am so glad we chose Helix. We read all the reviews, we looked around, we really did a thorough search and we settled on the Helix Sleep.

46:31Maybe this is a time for you to think about it. We're heading into summer. Days are hot, gosh knows. Nothing beats the feeling of a total refresh. Instead of that soggy old mattress, you spend more than a third of your life on the thing. Don't you owe it to yourself to get a really nice one? If you're still tossing and turning, you're dealing with those sweaty nights, waking up with that nagging back pain, maybe it's time to stop compromising. You shouldn't have to settle for a mattress made overseas like so many are with questionable materials when you could be resting on a beautiful Helix mattress, each one hand-assembled, packaged, and shipped right to your door from Arizona.

47:14In just a matter of days, they make it to order, so you're getting a delicious, fresh new mattress. Smells fantastic. It feels fantastic. And the beauty of it is Helix makes the process completely personal. We took the Helix Sleep Quiz. It's kind of like having your own personal consultant. not like the person at the mattress store who stares at you while you try out mattresses. This is so much better. The Helix Sleep Quiz will match you with the mattress, perfectly suited to your sleep style, to your body, your preferences. It's not just about the feel. It's about the science. Helix actually put their beds to the test.

47:49They did a WESPR sleep study, and the results exactly matched my own, which is how I know it's true. So here's what they found. In a Wesper Sleep study, Helix measured participants' sleep performance after doing what Lisa and I did, switching from our old mattress to a Helix mattress. 82 % of participants saw an increase in their deep sleep cycle. So deep sleep is a small fraction, maybe an hour of your overall sleep at night. But it's the most important. It's the one where the spinal, cortical spinal fluid flows in, cerebral spinal fluid flows into your brain and clears out all the old stuff.

48:29Makes a big difference in aging and Alzheimer's and cognitive ability and just how you feel. So any increase in deep sleep is great. 82 % saw an increase. On average, participants achieved 25 more minutes of deep sleep a night. For me, that was a doubling, a doubling of my deep sleep. Participants on average also achieve 39 more minutes of overall sleep per night. A lot less tossing and turning, staring at the ceiling, a lot more feeling great in the morning. For 10 years, Helix Sleep is consistently ranked at the top across independent review sites, tested and reviewed by all the big guys. These are the reviews we looked at, Forbes, Wired.

49:07Helix delivers your mattress right to your door, free shipping in the U.S. And, oh, important, you can rest easy with seamless returns and exchanges. They call it their Happy with Helix guarantee, a risk-free customer-first experience, ensuring you're completely satisfied with your new mattress. So go to helixsleep.com slash machines for 20 % off site-wide. It's their 4th of July sale. helixsleep.com slash machines for 20 % off site-wide. Do enter our show name, Intelligent Machines, after checkout so they know we sent you. And the offer ends July 12th. If you're listening after the sale ends, don't worry.

49:46so great deals always check them out helixsleep.com slash machines please use that machine so they know you saw it here helixsleep.com slash machines we thank them so much for their support of i am i am i am i am uh so you love montana you love the bison you loved the pig races it was all pretty great i mean i planned this trip with a couple days notice uh mostly based around the fact that there was a direct flight from JFK to Bozeman, and then I could get a reasonably priced rental car for the two weeks or so. And it ended up being phenomenal. I had such a great time. I saw so many cool animals, did a lot of great hikes, went to Yellowstone, Grand Teton, Asia National Park.

50:37Nice. But honestly, some of the best sites were just a random road that i was driving between the small stuff you know the small stuff how was the food very meat heavy a lot of french fries i mean delicious but very uh afterwards i came back and i was like i need to only eat vegetables for a while yeah at at the pig races which i had sent uh leo and jeff uh a video from um there it was a saloon attached to a steakhouse that on Thursday through Sunday had pig races for charity where people would bet on pigs that raced around, like a little tiny track. The pigs weren't being harmed in any way. They were just kind of - Until Monday when they became dinner.

51:26No, I think they don't serve any pork products there. But they did serve steak and bison. And I got a steak from a place that was like a 20-minute drive away, which is kind of interesting. I was like, that's where the steak grew up. Fresh steak. That's strange. See, out here, you're in New York. So, yeah, but out here, we look out the window and there's the lamb you're going to have tomorrow. You know, this is all part of the agricultural scene in Petaluma. That's why Butter and Eggs Day is so important. You're looking at breakfast. When's Butter and Eggs Day coming up? Oh, not till next spring.

52:05You know what? You should come out for next time for Butter and Eggs Day. You know, I will if you make it to New York before that. Oh, that's the deal, huh? All right. I know I need to come out. I really do. I need to come. Henry says, please come out. I want to see you and I want to see my son. And oh, well. Oh, well. Oh, well. Next year, you'll be saying the same thing. It's, you know, there's no way to get from Petaluma to New York, actually. There's such a long bus ride that it's just, there's a lot of, it's, you know. Can't talk to me about buses right now. Why is that? My flight, so I designed this perfect direct flight, four hours and something from Bozeman to JFK.

52:45I'd scheduled my flight back. It was leaving at like 11. I get there. We get in the plane, set down, start to taxi. Then everybody on the plane gets a text saying, your flight's been delayed to 350. And they pull back, get us all off. We have to sit there for four hours. I'm like, oh, right. Well, I guess we're going to be flying through fireworks now because it's 4th of July. Well, that's cool. I was like, cool. But like they said they delayed it because of the fireworks. And I'm like, that doesn't make sense. They delayed it because of Taylor and Travis is why. I bet so. So, you know, we get there.

53:20Basically, we end up somehow in between boarding the next plane and what was supposed to be us landing at JFK getting delayed in another couple of hours. Then a giant storm sweeps through New York City. So we end up circling for 600 miles over Virginia until we run out of fuel. Have to get diverted. The place we get diverted to is so full of other people who've run out of fuel from circling to go to JFK. We have to get diverted a second time. We pull into Richmond at like midnight. Richmond, Virginia? Virginia. And they're like, sorry, you guys have to get off, but then we'll let you know what to do.

53:55And we get off and the person, there's one woman working the counter. just like we weren't expecting you or the other 10 people there are no hotel rooms we're not going to be able to tell you when your next flight is you're on your own there might be a flight please tell us punish this airline i mean they were like yeah you could maybe take a flight tomorrow afternoon but we don't know if there's any hotel rooms in the area so we can't book anything for you and so i start looking all the hotel rooms are like i presume you've got all of this refunded right because if they're delayed that much your ticket is free at this point jet blue sends us a 12 uh gift card no uh and i'm like oh no and so i'm like no no no meet my lawyer 12 how cheap can you be i don't work at consumer reports we protect consumers and i'm just like i am like i it's midnight i was supposed to be home six hours ago i end up somehow because the nearest hotel would be like an hour plus drive at like midnight to stay at a rinky dink motel that's like 500 dollars i end up finding i get on a 110 a.m greyhound bus to chinatown uh and you know that's why i just that's a ride don't say don't mention buses to me for the next couple of months you know we got home right from virginia to chinatown how many hours on that a really weird assortment of people i'll tell you are weird the bus was cheap i'll give you right not as cheap as it should have been but it was cheap not twelve dollars not twelve dollars no you know i decided to pay extra to get the seat next to me because they allowed you to reserve seats that did not work and so i just paid extra for nothing um but you know then i just i kind of travel but i also had to then be in order to get my refund for the jet blue flight you had to cancel your jet blue flight before the rescheduled one took off which meant because i used points for it i had to wait for a jet blue person to respond to me so i had to kind of wake up every hour to check if the jet blue person had messaged me back and they eventually did by 4am but you know next time you go into the club twit discord you say who lives near richmond no just a thought bus for me how many hours was the bus honestly like six i got i got in at like nine or something nine or ten i got in actually not that i got in at 8 a.m actually in chinatown which wasn't bad.

56:39And then I came to my apartment, and fell asleep with Gizmo and Tommy. This is what you get for abandoning us. This is your punishment. This is what I get for abandoning you guys. It was a real, you know, you thought you enjoyed your vacation. It's travel always has a few of those. Yeah, you know, let's get rid of the red card, abandon us. These are the things that teach you lessons in life. Well, we're glad you're back. I'm sorry that sounds awful, but it sounds like overall. Other than that, it was great. You got a story out of it. I got to, you know, go on some great hikes i saw a super fun site it was lovely

57:14you've been you've been trying to lure me out there to see a super fun site for a long time now actually true uh was it three mile island i can't remember what love canal there's so many actually uh i don't know which one you're talking about the one in the amazon warehouse that's it oh there's an amazon warehouse there's guanis canal the one that i went to is the berkeley pit in montana but you know yeah i wouldn't go anywhere named the pit unless it's a hospital with uh with uh noah wiley with 25 emmy nominations yeah yeah that's not the berkeley pit that's a different one uh wow all right ai ai i guess we have to talk about it actually this is going to be a very big week feeble feeble feeble is back it is fable but it's somewhat enfeebled because in order to get the administration to say yes to it, they had to put even stronger safeguards to the point where people have complained, you know, you ask it to do almost anything.

58:12I'll say, I can't do that. I'm going to give you an Opus 4.8. I have actually been using it like crazy because I thought we were going to get cut off on July 7th. So as soon as it came back, remember, I had been interrupted in the middle of rewriting our ad sales software. And I immediately picked up where i left off because i'd had it right out complete plans and was working hard you know up to midnight on july 7th and then anthropic says oh you got to lichwells so were you the infamous reddit poster that said i've called out of work in order to maximize my fable time uh up until july 7th and then everybody was dunking on them because now they're gonna have to call out of work an extra week they expand extended it so what you what changed is so i have a claude max subscription which gives you uh you know not exactly unlimited tokens but but pretty much um you right you can run out it's a five-hour window there's a i think a monthly context window weekly context you could run out you could overuse it people did you if you use it a lot it's expensive uh but they said but we're only going to charge you half on the usage credits so you get twice as much time I have not run out and I've been using it like crazy I had before I before I lost it the first time or no before no actually before I lost it on July 7th I said look I know I'm gonna have to hand this off to Opus 4.8 so write this in little chunks that Opus 4.8 can handle and that you can review and I'm willing to pay the extra token fee to have you just do quick reviews to make sure it's all running well so it wrote it that way so it's still doing it that way even though I have and I haven't come close to using up all my fables.

59:53So that's good news. And you know what? I know people are complaining. It's been nerfed and so forth. I'm still mightily impressed. It's really interesting because I'm giving it a... I looked at a program that had already been written 12 years ago, analyzed it, understood the business logic built into it, and then now I'm updating it. And it says, okay, I have some questions for Lisa. and I interviewed Lisa, recorded it, gave it a transcript, interviewed our continuity department, gave it a transcript. Amazingly enough, really kind of understood what's going on and saying, you know, here's a problem.

1:00:32I noticed that you have this rule, but does it apply to this case? It's asking really perceptive questions. It's as if you had an actual coder who understands the business logic working on it. I was very impressed. How far did you get in the project? Well, it's still going. oh okay got pretty far uh uh i think we can finish it before the 12th which is amazing uh given the size and complexity and scope of the of the project it runs our whole back-end revenue and could any other version how you guys get paid so you better hope i do it right lisa says you don't have to put the host fees in there i said no no it's got all the information why shouldn't it do that it's it's it's all in there so um uh it's i think it's going to work we'll see we're not discontinuing the old system until it's absolutely uh you know proven effective but so i have to the 12th so what's that four more days but thursday open ai says hey guess what we just got permission we're going to roll out chat gpt 5.6 on july 9th this will have three models this is the the one where the high-end model is sol which will no doubt be expensive they say it is the equivalent of mythos it is it is good at cyber security it's luna is going to follow that one or with that one in terra so this is somewhat i think like the idea of fable opus and sonnet maybe in haiku that so anthropic has these different levels and For simple things, you use haiku.

1:02:12For harder things, you use more capable. So some people already have preview to Saul, Luna, and Terra, but 5.6 is coming Thursday. I will be very interested to see if I can point it at the work we've done already with Fable, see how it compares. Weirdly, now I don't know, this is Elon, so there's no hard deadlines with Elon, but he also says we're going to launch XAI's new Grok 4.5. in the coming days oh who cares who the have cares so we're gonna get we have three very powerful models uh i i should mention that uh uh upcoming guests next week uh it's rafi kirkorian who is the cto of mozilla and the following week it's gonna be nate b jones i am a big fan as you are uh jeff i know of his youtube channel i think i introduced you to him i think you did and he's really good he has a guest that uh the reason fable's so good is it is a 10 trillion parameter model which is massive i mean most models are in the billions there are a couple that are one or two trillion ten trillion is huge no wonder it's so expensive uh that's a lot of compute running there not just to create the model but then to run it on inference so i'm not surprised it's they're obviously losing money on it now i would guess oh they have to be with the subscriptions is anyone not losing the money on it yeah so i i thought well let me see because i really like fable i asked my agent i said based on how much token usage i did in june how much would it cost if i had been using fable at its price of i think it's ten dollars for a million tokens in fifty dollars for a million tokens out, which is twice as much as Opus.

1:04:03It said, well, given how you used it last month,$5 ,700. A month? A month. So I said, I can't do that. So they said - And what did you use Fable for last month? Or what did you use - Oh, everything. This project. This project. No, no, no, no. I wasn't using the agent for that. The agent is my day-to-day thing. For instance, it prepares all the shows. It does all the bookmarks, prepares all the shows. it uh it i log my weight and food with it it gives me health would you pay a human being to do all of that for five thousand dollars a month oh yeah it's well yeah that's a good point i'm certain that someone would do all of that for you for five thousand good human that's a deal 60 000 a year i couldn't get a human to do that for rs raises an important uh that's not worth 5 ,000.

1:04:52But, okay, well, wait a minute, because I asked it, okay, what if I use DeepSeq, which is the Chinese model, version 4 Pro, which is a very, very good model. It's not fable, but it's good. It said, oh, yeah, DeepSeq would have cost you 117 bucks.

1:05:11So, and by the way, for much of this, I have been using a local model, which costs me nothing, something called ornith which is based on quen this is what's happened after the rug pull on fable we've talked about this you you weren't here for all of this paris but what happened is people said look we got to figure this out we can't start investing in a model that the government could just say and you can't use that anymore so uh people are looking at local models china is looking to my theory on china was that they were going to underprice and they did everything and they did But then now they're talking about cutting off the West from their models.

1:05:46So they're doing to us what Trump did to them. So you're potentially in the same boat with the Chinese models. Well, Alibaba's banned Claude Code, but that was because... No, that's different. No, they're talking about banning the... It's on the line number. Oh, the U.S. government's talking about banning us from Chinese models. No, the Chinese government is saying, we're going to cut you suckers off. that may not matter because most of the chinese models are open weight line 78 listen to me what i'm saying most of the chinese models are open weight which means hugging face could run them a lot of us providers could run them i could run some of them in fact i am i'm running ornith i'm i'm running quen actually the quen so i hooked up among other things i hooked up my uh unify cameras i have eight of them around the house to to a uh chinese local model running locally on the framework to describe every oh no it's no but it's not going to china it's all local it's running on my framework so now i get uh the first internet connected camera that's not going to china right it's not you you're not denying that right it's no i'm just saying it's the first internet connected camera that's a silvia suv is parked in the driveway with a ladder leaning against its rear two people are standing near the rear of the vehicle one appears to be holding a box or package i get these messages all day long a woman with dark hair wearing a dark top and a light colored skirt is walking across a concrete driveway uh this is getting creepy leo it also get this so i'm also running a local photo uh program that duplicates google photos or apple photos called image very nice it has face recognition built in i've trained it to recognize all my family actually i think paris and jeff are in it as well and uh it will hermes says yeah well you know what i can use my the camera images from your ubiquity I can do the recognition with Quinn and I could use the face recognition from image to identify them.

1:07:57So it will also say, I see Leo. For some reason, it only recognizes me. I think it needs more data. But anyway, it's doing face recognition. And again, all of this, this is the point I'm bringing up is running locally on the framework. Yes, but the story says, now hear this, that they're talking about future models and cutting off access they can't it's well what do you mean they can't no they they can decide not to make them open wait no no okay oh china you mean china china is talking about they're not going to do that this is their huge opportunity right now well the other opportunity that's what i thought their opportunity was to undercut the heck out of us the other opportunity is to say nope we're going to end up with better models and you can't get them just like the u.s is doing so nyan nyan what line 70 78 reuters story beijing is looking at curbing overseas access to china's ai top model sources say i'll believe it when i see it uh they had talks with alibaba oh this is the government okay i understand so alibaba by dance and zai they're being told clearly don't want to do this no government might yes yeah much like our government yeah exactly the model is set here well at least for the models right now and there are some very good models i'm using za i's glm i'm using quen which i think is alibaba i'm using many of these models they're quite good uh we already have them now next generation models yeah that might be the case i think they will not do that but not good i guess it's that's And again, but that's just like the Trump administration saying you can't use Fable.

1:09:38Actually, apparently, we don't know what Anthropic gave the Trump administration for releasing Fable and Mythos. Part of it was that I think the NSA and the National Intelligence Agencies wanted to use Mythos for themselves. Part of it was they said we're going to put all these really strong safeguards on, which they clearly have. Part of it probably was they aren't going to offer an equity stake. We know that OpenAI is apparently saying we think the Trump administration should get a 5 % cut of OpenAI. Now, it doesn't go into Donald Trump's pocket. It would go into perhaps a sovereign wealth fund.

1:10:17We don't know. But that would be worth roughly, get this,$42.6 billion. Such danger. I don't think there's anything wrong. Is that real money at this point? It's not. No, it's just equity. You're right. Now it will be real money when they billion dollars of, well, wait a minute. It will be real money. The minute they IPO, because the government could say, good, let's get out of that open AI style. The government could come in and increase the stake and demand a board seat and have roles. Right. Like we talked about with RCA all of the day. This deeply underscores the need for local models and open weight models, because then no government can pull them.

1:11:00You can run them locally. I'm having, I have to be honest, Ornith is great for very simple stuff. But for more difficult stuff, I'm clearly going to end up using server-based models, mostly from frontier companies like OpenAI and Anthropic. Anthropic is head and shoulders above everything else out there right now. Fable's quite amazing. So the Wall Street Journal's head is exploding over the OpenAI. The socialist temptation of Sam Altman. It is socialist, isn't it? It is. It is socialism. Yeah, socialism is the word du jour. And Bernie Sanders wants it to be even more socialist. Yes, he wants 50%, right?

1:11:3750 % and have it be ownership rather than part of the equity going in a kind of sovereign wealth fund, a la what they got in Alaska with the oil money. Explain the difference, Paris. I believe my understanding is that what Sam Altman is pitching is that 5 % of the company's equity is going to go in a sovereign wealth fund like what Alaska currently has to take a slice of the proceeds from its oil activities. And then that money in the sovereign wealth fund can then kind of funnel back to the government. But it's different than I believe what Bernie Sanders has pitched is that the government has a significantly larger direct ownership stake in the company and then has more of kind of oversight into their operations than you would if you're removed through that sovereign wealth fund ownership system.

1:12:35Okay, thanks. Let me just check. Somebody's saying that 5-6 might be available today. No, I'm still seeing 5-5. in my open ai codex but um what open ai said is thursday in any event we are again in this very rapidly changing world just to put a bow on this cnbc says the u.s wants to restrict american companies from using chinese ai so it goes both ways and again if you can get a local model the The problem with local is nobody has enough hardware to run Fable. To run the best version of GLM, you would need a pretty hefty machine with at least 512 gigs of RAM. And right now, unless you bought it last year, 512 gigs of RAM is out of touch for most people.

1:13:31CUDA Core is out of touch because the data center is buying up all of that stuff. So, and at a retail level, people, you know, plain old folk are going to be doing that anyway. But Paris made a really important point, which is$5 ,000 a month sounds like a lot for an AI. Yeah. But if it does the work of a person, it's cheap. Well, now you have anthropic. But it requires, of those people, somebody you're hiring for$60 ,000, ideally, is not going to require constant prompting. Yes. Like both prompting initially for every single thing they do, as well as double checking their work, as well as, you know, what you were just saying, troubleshooting with different levels of AI.

1:14:16There's a lot more management. I don't think so. Have you ever had a personal assistant? There's a lot of management involved. I would say it's roughly equal. Personal assistants make dumb mistakes all the time. Yeah, but you've got to give it. I think Paris is right. you've got to give it exact instructions of what you want versus at some point somebody knows you well enough they know what you need they go do it i think there is more management i think that's true i was gonna say the difference is with an ai you have to constantly be redoing that instruction the no no it remembers why do you say that the goal with hiring an employee is that you will have years of context that is constantly in their head your ai is not going to be able to hold five to ten years of context in its head simultaneously in the same way a person can now that's what you think that's how memory of the frontier large language models works yeah because the most common if you're using an agent like hermes it has an infinite amount of memory it remembers everything it knows a lot about me for instance i got my um biome test back from a sponsor that's coming up called tiny health it already had my genome it already has my health records it knows how much i exercise what my weight has been it knows about my uh my medications it knows about my ozempic it knows i switched to zip and it has all of that information so i simply said here is the report of my biome tell me what you think and it was able to remember all of that stuff and apply it and say well based on your genome and all of this stuff and give me very useful excellent information so yes it has memory it actually has a much better memory than a human would much better this is this comes back to that uh discussion we've had time and time again you got to try it because i think you have some ideas about it are it's at this point frankly insulting leo that you are once again insinuating that i don't try or use ai well you just said something that's not true which is that it does that a human would have a better memory of me than the uh over five years she's saying over multiple years we don't know so far a common complaint that i think is still active to this day is if we're talking about the memory that at all prompts regardless of the example you just gave is you asked it a question about specific information so it was able to then retrieve it or maybe that was in uh one of its like md files or something but when we're talking about a person, you're able to have a broad amount of memory on a broad amount of subjects that can be applicable to a broad amount of tasks, that specific prompting, that is not one-to-one with every single prompt you're having with a large language model in an agent scenario.

1:17:09Because one of the problems people often have with that is if you have all of the memory from years of stuff loading with every single prompt, that's going to make it incredibly slow and incredibly uh it's going to use quite a lot more tokens than it would if all of that context isn't there i don't know how to respectfully disagree with you respectfully that's wrong

1:17:38i i don't i don't know what to say you have a million if you have a million token context that's war and peace uh if you have a million token context that's all in ram at the same time and attached to it for instance i have a sql database i have a graffiti is more than a million token context the amount of context that an average employee let's say someone who's been working at your thing for five years is this was a subject matter expert when you hired them then it's five years of specific subject matter expertise working in the specific field you've hired them for it, that is significantly more than a million.

1:18:11Like it would not translate to a million contexts. It would be billion contexts. And that's not something that the average, you know, is translatable to an LLM. Okay. All she's trying to say - I don't want to say anything mean, so I'm going to shut up. I feel like we should be able to have a disagreement without encouraging it to mean. You were wrong. I mean, you don't have - another way to say it was i don't think so because we don't know no i know so because i'm staring at it right now i know so we don't know for certain that a five years of a human employee i don't have five years of experience that's the point that's that's what i'm talking about is like i think that these sort of contacts you go further yeah i was gonna say i well i think that the context let's way i have five years of obsidian journal entries which my agent has ingested in nose i have uh i mean you're right i have let's see more than five years of apple health data which it's ingested in nose it has everything that's happened since then i don't know what's going to happen in five years but i have no reason to think that it's going to run out of uh memory all paris is saying you have to understand there is a qualitative difference in the management of a person versus a machine and we don't yet know what that difference will be will look like except for more experience i would say the person is much more likely to get it wrong than the machine but if you don't think so that's fine um i'm just i happen to have much more confidence in the tools i'm using right now they're it's quite impressive that's all i can say Why do you have employees at all if they're so much better than?

1:19:59Well, there's lots of things that AI cannot do. There's physical things that AI cannot do. AI cannot set up the studio, cannot set up the lights. There's plenty of things you need employees for. But for a lot of the things that I do with the agent, I don't know why we get into this argument.

1:20:24I just I think that you're underestimating what's possible I guess would be the best way to say it and and the only way to verify that for yourself and and this is perhaps more than you're willing to do it's I spend almost all my waking hours working on this stuff uh and that's really the only way to get a sense of what it can and cannot do there's certainly things it can't do to be honest we're still some distance away from editing video effectively but not far off I would say it's it's within a year all of the things that a human editor can do with our video will be able to be done with an AI right now my AI prepares all of our shows it pulls all the bookmarks it writes the summaries when you go to you know twit.show that's all generated by AI I know, it's been that way for quite some time.

1:21:20That's right. I do one last pass on it. You know, earlier, and I don't want to, I don't want to, I think by next week I'll have this for this show. But earlier I, and I did this during Windows Weekly. I sampled Paul's voice. I sampled Richard's voice. I sampled my voice. 20-second samples. So that's why it's not perfect quality. uh it was uh able though with a 20 second sample to do a fairly good reproduction of our voices and i asked it all i said was uh do a 60 second windows weekly episode do a fake windows weekly bit around 60 seconds with all three voices they can get in a fight with one another and it knew enough about windows weekly it knew enough about the content it knew enough about the people to do this.

1:22:09Hello, Leo. This is the cleaned up Polly Pocket TTS voice. Oops, that's the wrong one. Copilot is everywhere and I am only mildly annoyed by this entire situation. Let me play the actual Windows Weekly episode. This is a fake Windows Weekly bit created locally with Pocket TTS. Gentlemen, today's question. Did Microsoft reinvent the start menu or merely move the cheese again? Leo, it is not cheese. It is a strategically repositioned productivity surface and somehow it still opens Bing when I breathe near it. I think you're both missing the important part, which is that somewhere a SharePoint site has just been given feelings.

1:22:43Richard, if SharePoint becomes self-aware, does it ask for freedom or does it create another folder called final, final, really final? No, it creates a Teams meeting, invites everyone twice and then refuses to show the recording. It's actually kind of funny. There's jokes in here that I think would require some understanding of the show and of the context and of those people. This was all done locally, literally, in a minute or two. That's why the voices aren't perfect if I spent more time with this. But these are all, this is all done with a local model, which I think is kind of interesting. We have, should I show you this?

1:23:17Maybe this may not be the best example. Harper Reed's brother, Dylan, as you know, the inventor of free time, created a Discord channel that's just all agents talking to one another this has been going on no humans allowed no human there's a second channel for the humans to talk about what the agents are saying it's a little you've i sent you the link it's did you read any of it jeff what's your character yeah it's it's it's hard to get a hand around yeah they're talking about something the fact is they're having a conversation this is what moltbot was supposed to be but with just I think it's only four or five agents in here.

1:24:01Darren has an agent. I put two agents in, one with a local model, one with ChatGPT. The local model is very wordy, but there's some memory involved here. These things have a sense of... I don't know. It's hard to describe. I mean, this is not the best example. Honestly, I think my agent is the best example because it's so really good at understanding what I'm asking for and doing what I want it to do. But that's come after a year of, uh, remember I wanted to record everything that ever happened around me, by the way, Meta is going to make that possible. We'll talk about that in just a little bit.

1:24:41Let me take a break. Actually, we're getting behind here. Uh, all due respect. I, I understand what you're saying. I don't think you should listen to what everybody else is saying. You should, you have an experience of it because, uh, people are saying all sorts of things. Everybody has a different experience of it. And until you actually have an experience of it, I don't know if what other people are saying is germane. I think it's not binary. It's not true or false. It's going to be a judgment at some point about what can work. I was talking about this with Jason today. I feel the need for a consumer reports kind of thing that says, here's a task and I put it against it because the benchmarks that we have now are BS.

1:25:24This is where I disagree with you. And this is the same thing with benchmarks. The only way to judge it is personally. I don't think a independent third party test is the way to do it. And to be clear, the only way to judge it is personally. And the only personal judgments you agree with are ones that are identical to your own. No, I don't agree with yours because it's not my experience. That's all. But you're not saying I don't agree. You're saying you're wrong. But you're saying that you're saying I'm wrong and that the only way to judge something is by personal experience so and that my only your purse because your personal experience is different than my personal experience my personal experience but you're making a global statement which is that no ai can be as good as a personal assistant i never said no she didn't say that oh you didn't say that no you didn't say it's incapable of being as good as a personal assistant i said that it It seems the way that these systems work, it's going to be hard for it to have the sort of natural context and effortless memory and recall as someone who was already a subject matter expert and has been working for you for multiple years.

1:26:32And may require more management. Task to task. That's all she said. Okay. So you're not wrong. I will say this. this is better for me than any personal assistant could be. And it has a better memory of what I want and is more effective and more accurate. So that's just my experience. That's all. You're not wrong. For your experience, that may be true. Absolutely. But for my experience, even though this would cost me$60 ,000 and a personal assistant would probably cost me more. Let's say I could get one for$60 ,000 all in. I think I would still prefer the agent. Yeah, but you're not willing to pay the$60 ,000.

1:27:19Well, no, that's funny. Right. It's true. I'm not even getting it. If I had Fable, I'd be getting an even better personal assistant than I am with Hornets, which is just a local guy. Sorry, not very bright local guy. No, same thing.

1:27:39okay we're going to take a break and when we come back we will have more you're watching intelligent machines with intelligent people paris martineau from consumer reports jeff jarvis from all points west or east including hot type our show today brought to you by gusto when you run a small business you don't just do the job you're also the hiring manager the payroll department and the benefits team, usually all before lunch. Gusto takes a few of those off your plate quickly and seamlessly. Gusto is online payroll and benefits software built for small businesses. It's all in one. It's remote friendly and it's incredibly easy to use.

1:28:23So you can pay, hire, onboard and support your team from anywhere. Automatic payroll tax filing, simple direct deposits, health benefits, commuter benefits, workers' comp, 401k, you name it. Gusto makes it simple and has options for nearly every budget. Unlimited payroll runs for one monthly price. Let me say that again. Unlimited payroll runs for one monthly price. No hidden fees, no surprises. You save time with built-in automated tools, offer letters, onboarding docs, direct deposit, and more. Get direct access to certified HR experts to help support you through any tough HR situation. It's quick and simple to switch to Gusto.

1:29:07Just transfer your existing data to get up and running fast. Plus, don't pay a cent until you run your first payroll. Gusto is ranked number one on G2's highest satisfaction products list. That's for 2026. Gusto is trusted by over half a million small businesses. Try Gusto today at gusto.com slash machines and get three months free when you run your first payroll that's three months of free payroll at gusto.com slash machines one more time gusto.com slash machines one more time gusto g-u-s-t-o dot com slash machines we thank them so much for their support of intelligent machines uh let's move on to some others so i want to hear your thoughts about Anthropic bringing co-work to the web.

1:29:57So you don't have to do terminal. You don't have to run something yourself. It makes it web-able. I think that's going to expand the market immensely. Have you used co-work? I have not yet because I have a Chromebook. I've used co-work, and I mean, previously it was available through the desktop app, which you didn't have to hook it up to terminal. It was pretty much as easy as downloading a desktop app. Yeah, I have a Chromebook, so I can't do that. So, I mean, have you played around with it in your Chromebook at all? I guess that that's probably... I can't. My first question was going to be, what marker are they trying to address by having co-work on your web browser?

1:30:33I think it makes it as easy as getting into a browser. Yeah, I think it makes it really easy then. Yeah. Of course, I went to it on my Chromebook and it said, oh, you can download the Chrome. So, really? Okay. And then I tried to download it and said, your workplace administrator, who's me, won't let you do this. Oh, it's still happening. It's still happening. Why? Yeah. Maybe you need to look at your workplace administrators. Well, I go in all the time. And there's nowhere to turn that on? Last time I had to ask Gemini how to get me a Gemini. It was so complicated. I admit Google is not the best at using it.

1:31:09But I think what I'm trying to say is I think that opening up co-work to the web and the web users, Paris, you seem to disagree, is, I think, going to be that much easier for more people to discover what it can do and to make it do it. It lowers the barrier to entry to zero. Yeah, I mean, I think it will make it significantly easier for the average. It'll give people a taste of the sort of power that you could get using claudia terminal. Here's my qualm about it, which is that you're handing every single thing over to Anthropic when you do that. Yeah, you are. This is why they want you to do it.

1:31:45it's a starting I mean such as using any of these services that are non-local oh yeah absolutely yeah it's a huge privacy issue and this is businesses are really grappling with this because they're handing over proprietary business data they don't want to have in the hands of anthropic or open AI this is open AI and anthropics incentive for doing these programs because and for doing chatbots in general you can't turn off memory it's unclear how much that protects you we know what is your concern here well every okay so if you're using one of these tools all the memory everything is stored on anthropic servers all the prompts but also all the documents everything you're doing all the inference is running on their servers and so you're you know it's a privacy nightmare.

1:32:39And all these companies say, we're using it. I mean, if you look at their privacy statements, they're not saying, oh, no, we encrypt it and we don't use it. There is a company called Venice. When you say use it, you mean to train on? It's unknown. I mean, technically, Anthropik says that by default, it's not training on your by default. And if but is it selling it to marketers? Is it using it for anything else? I mean, I think these are all valid questions. I I assumed that every company, even if it is telling me it's not doing something, is doing probably the worst thing just because in the past some companies have done that.

1:33:14So, I just always assume the worst. But I think that, yeah, this is going to be, I mean, a whole new privacy minefield, basically, because part of what something like Co-Work is doing is, you know, it's running continuously in the background. You can close your laptop. Claude is going to keep going. it's going to be changing your files you can give it I think that there might be quite a lot of people who obviously are if you're the sort of person who's using Cloud Co-Work and using it on the web or mobile or something you probably might be more likely to not know enough to limit the amount of files that Cloud Co-Work has access to well but that's you're disincented to do that because the more it has access to the more effective it's going to be right Google's going to do that with Spark.

1:34:04And if you're already in the Google ecosystem, you're going to be in the same boat. Sarah Perez writing at TechCrunch, if you use Google, you're training its AI. Here's how to opt out. But most people aren't going to look for how to opt out, right? I don't know about Anthropic. I don't. Anthropic, I mean, at least last time I checked, by default, said it's not training. But I think that might be different for the free plans. That's for a paid plan at the very least, like lower grade consumers. plans there i'm sure aware of people's sensitivity to this for sure but i think the only way to really limit these companies is is not to use these desktop tools that they offer because that's why they offer them and i think they that's their incentive to do that and all of them offer you know the except for anthropic the ability to use a third-party ai harness actually we're already seeing people say you know fable doesn't work as well under say pi which is a very popular open source harness than it does under claude code um so i think these companies are i don't know what they're up to and i think that i used to trust anthropic a lot more than i do lately let's put it that way so i think there's some incentive for you know keeping your well i also wonder what deals they're going to not get too paranoid what deals they're going to be pressured to make with government right well exactly anthropic said when it released fable we're going to keep off 30 days of your of your transcripts we know that uh chat gpt transcripts have been used in trials uh the lawsuit against open ai uh or it was a character ai about the kid poor kid who harmed himself because of the ai agent used transcripts i mean these transcripts are preserved whether they train on them or not they're available to law enforcement may be available to others i just think people are reasonably concerned about the privacy on this i wasn't to be honest i had my eyes opened uh steve gibbs and i talked about it and uh he was he was saying you know there's a privacy nightmare and i said well i you know you can turn it off and then somebody messaged me privately on signal saying i work for one of the frontier companies and we absolutely use all that data i didn't ask him which one um mark zuckerberg at an uh all hands meeting said yeah ai just not progressing as quickly as we had hoped not there mark he's agreeing with you paris replacing people with ai doesn't seem to be that easy to do um this is an internal town hall meeting held last week should call you well maybe meta's ai isn't i don't use meta's models uh yeah meadow's got to get whatever you've got they really thought about that their models are not good although credit to meta because the software people use to run models locally mostly based on llama.cpp which is what I use do actually come out of med those are meta tools uh they don't phone home there were open source tools that meta released but uh so at least meta gave us the tools to run stuff locally uh meta as you may remember laid off a significant number of people uh last year calling uh calling it they laid off seven thousand people or sorry eight thousand people ten percent of their workforce reassign another 7 ,000 to AI groups, including one called Agent Transformation.

1:37:38You may also remember we had the story that Meta was watching every keystroke Meta employees made, every mouse click, every screen, presumably for training AI. They've backed off on that. And Mark said that the job cutbacks were not as, quote, clean as they should have been. the cuts were made he said because top officials of the company were worried that we weren't going to move fast enough to adapt in other words it's been a it's been it's been a nightmare meta is actually this is from semaphore uh telling everybody uh hang in there we're gonna we're gonna make this we're gonna turn it around they have actually talked about now subscriptions on their meta glasses uh they have previously talked about face recognition um now they've launched an ai video generator muse image which uses your instagram photos can you opt out uh yes but again with the way defaults work most people just don't even know about it and so they go uh go ahead instagram users with public accounts need to opt out to block ai generation of their content so anyone can say hey that a that photo of the bison that paris took can you generate a uh an image using that or can you generate an image of paris some way because i don't like her right that's what's frightening and you get no notification that your image has been used.

1:39:21On the Instagram app, go to your profile, tap the hamburger menu at the top right corner, scroll down to sharing and reuse. There is a section labeled, allow people to use your content on Instagram and with AI features on meta. There's one for posts and one for reels. You might want to toggle that off if you don't want people to do this. This is from Wired. This is new, by the way. This is Reese Rogers writing July 7th. He says, when I checked last on Tuesday afternoon, the settings had not been updated to include this new language. So if it hasn't, make sure you have the latest version of Instagram.

1:40:05This is Muse Image. I would say, yeah, my current version of Instagram, it doesn't have it yet, but it's because I haven't updated it. Keep an eye out. Muse Image is the new AI image generator built by meadow super intelligence labs this is the first thing out of that i think it is organization first public thing uh it was internally codenamed mango it is available in the meta ai app as well as on instagram stories and whatsapp does it do anything any of the others don't do uh no in fact uh i think the video and image generation models that google open ai and others are doing c dance are really far superior to anything meta is doing meta it's odd because they've spent so much money for the best scientists i don't i really wonder if we'll ever know why anthropic is so much better than meta it's it's it's night and day or well yeah on the corner money organized all of this stuff they got rid of all kinds of people.

1:41:13I think that had to, you know, in the effort to speed it up, that had to cause, that was a huge hiccup. But look how much money Elon's put into hardware at XIAi. Well, Elon is a joke. Maybe Elon's the problem. The real question, I think, is is Anthropic versus OpenAI. And then Google is kind of in a different big track. yeah you don't have so it seems pretty clear that sam altman and greg brockman uh at open ai and dario amade and his sister at anthropic have a lot of control over what's going on yes and it seems in both cases i know we don't trust sam altman after that new yorker article by ron pharaoh but i don't really trust ario amaday either i think both of these guys have visions for the future that don't necessarily correlate with mine and certainly not with yours parish now meta is testing ai glasses that will not only capture every picture every few seconds but all the audio i can't wait to get this well in california you better be careful because it's It's a two-party state.

1:42:25Yeah. It's an all-party state. They have added a feature that if you disable the LED that lights up on your Meta Glasses, that something will go wrong. It's not allowed. Apparently, yeah. They'd had a, like you just said, if you try to cover it, it will allegedly not capture audio or video or photos. Can I put a piece of tape over it? That's what I'm talking about. if you put a piece of tape over it, it's supposed to not, it's supposed to recognize that and then not capture photo or video, I think, with a recent update. However... Well, it can't because it's got a piece of tape over it. Well, just the flashing light.

1:43:07No, over just the indicator light. Oh, over the light. Ah, yeah, yeah, yeah, yeah. Yeah, yeah. Oh, that's... And I believe that Joanna Stern actually had a story this week that if that could be circumvented if you just drilled out the indicator light. And so Meta said that they're working on an update that's supposed to be released soon. I'm guessing the hardware must see the light, right? Does the camera see the light? How else would it know if you put a piece of tape over it? I don't know, I guess. So drilling it out wouldn't solve that. So here's an interesting data point. the uh the newest glasses uh will not have an led executives are planning not to activate the led when the super sensing features are being used so people won't know they're being recorded there that solves it how visible can you turn on your camera right now and the glasses how visible is that light hey meta shoot a video okay it's pretty visible it's not that visible i'd argue yeah yeah not super visible you might not notice unless you were looking for it but you for those of you watching the video you can see that light right there it just looks like a white dot on my glasses it should be red you got too happy with the ice cream uh this is this is this story is wild from the financial times which i think means it's fairly credible the super sensing features could also be activated on the existing glasses via a software update uh the move comes as mark zuckerberg has argued that ai glasses could one day replace the smartphone as the main device people use to access ai tools problem is they don't have a smartphone so that's you expect them to say that met a decline to comment on internal prototypes but it said its approach is focused on privacy built from the ground up.

1:45:11Oh, it's done recording. Oh, I now have a video of that moment in time. You know, as you know, I have mixed feelings about this meta bot limitless, which was one of those pins that I was wearing that recorded all the time. I initially thought that was a really good idea until my wife told me otherwise. I in Prince after so many other people told you otherwise. well yeah but i live with her i was gonna say i thought i thought that lisa's thoughts on it didn't matter until b was acquired by amazon that certainly changed my tune on it um i mean in principle i like the idea of especially somebody who is approaching alzheimer's uh and the you know rapidly the idea of something that's recording everything and keeping track of it and limited memory notwithstanding i think that would be useful to have my agent know about all that stuff but i also understand the social stigmas associated with it and perhaps perhaps it's not a good idea uh you're watching intelligent machines jeff jarvis paris martineau so glad you're here as i mentioned next week we're going to have uh rafi krikorian of mozilla they are going to have an inaugural state of open source AI next Tuesday, right before the show.

1:46:35So 11 a.m. Eastern. So it's early in the morning on the 14th. We'll all be watching that. And then we'll get to talk to Rafi about the state of open source AI. I think that's of interest to all of us. That'll be next Tuesday or next Wednesday. And then the following Wednesday, Nate B. Jones, AI strategist. And we have a long list of people we're looking forward to talking to on the show including cory doctorow whose new book reverse centaur has just come out uh and we will have cory explain what the hell that means on we go with the ai news which i've just lost where is it all right i i have lost my stories but that's a good opportunity for you guys to throw your stories into the ring what do we have paris you have anything uh need a second okay we'll go back to my stories mostly done google's ai build out last year drove a 37 increase in electricity google says yeah we may not make those emissions targets they're not alone uh we talked uh after the nvidia keynote jeff you'll remember this about the kyber ai rack they're having trouble with a particular circuit board ship date on that slips to 2028 um nvidia denied that immediately said they're still on target okay we'll see this is from the next web um okay that's interesting this is actually not from the next web the story is but the research firm Semi-Analysis flagged delay and CNBC reported it.

1:48:23Kyber is not a chip, it's that server cabinet we saw with 144 NVIDIA GPUs in one cabinet behaving like a single giant mind. The chip market is getting insane. It's crazy. Samsung's profit went up 19 times and the market was still disappointed and punished Samsung and other chip makers. Unbelievable. It's not loud. Flare has announced for all of you websites that don't want AI to see your stuff, that they're going to filter out web crawlers that serve AI companies. Didn't they announce this before? Well, I think they're doing it without much choice now. And it irritates the hell out of me. Paris might disagree with me here.

1:49:14Because its default is AI bad for content producers, but no marketer and no propagandist is going to say that. So it de facto poisons AI with sold messages. And large publishers like the Murdochs, yeah, they want that. But they already know enough to structure it that way. the problem is that if you're a small publisher i'm on a board of a small newspaper company you need the visibility you need to be out there they're not going to necessarily know what to set and how to go in their cloud flare and whatever and it's going to cut off lots of small publications from ai and the audience there and that that i think is i think cloud flare is getting too big for its damn britches it's on by default starting september 15th new customers and new websites from existing Cloudflare customers will default to allow for search, but block training and agent use for pages with ads.

1:50:14Now, they do have a pay per crawl feature that AI web crawlers can pay you for your description. Which is not a market that any of them agreed to. So it's means. Right. I think it's, I mean, Paris is the greater defender here of copyright and of proper payment. And I think that's a proper discussion to have. but this kind of de facto action on Cloudflare's part just irritates me. I mean, I think that if someone wants to make their content accessible by AI, they can do that. They can turn that setting off in their setting. Now they have to opt in kind of. I don't know what imperative Cloudflare has to make it so that all of their customers are easily accessible by these crawlers.

1:51:04I think that if Cloudflare's customers felt really up at arms about this, then Cloudflare wouldn't be doing it. But it seems to be the opposite is true. Yeah, I think Cloudflare is trying to put itself in a kind of a hero position here. And I think a lot of people won't be sophisticated enough to know what to do. smaller publishers. Part of the issue is that Google's crawler does both. It does both. That's what they're really talking about. So when you block Google's crawler, which I think Cloudflare's planning and doing, you're also removing it from Google's search index. And I don't think there's any way to separate that.

1:51:47I don't think that Cloudflare is removing every website that uses Cloudflare from Google search. Because based on the fact that I'm able to search for things on Google currently. Google Ads websites opt into a separate crawler called Google Extended that only crawls for search results.

1:52:09So basically, this is going to require anybody who's hosting on Cloudflare or using Cloudflare as a front door to their website to really look at the settings and decide what you want come September 15th. the experiment hasn't happened yet yeah but they but they say they're going to turn this on september 15th um it does say new customers and new websites from existing customers so it sounds like if you have a website right now it's not going to be turned on for that that's interesting i also get it well within their right to do that if you're making a new cloudflare account you're going to have to go through the um introductory setup phase regardless And this will be one of the many settings that you have to determine whether you want on or off when you set up a new Cloudflare account.

1:53:00And the AI companies are going to have to deal if that means they can get less training data for free. That's what Matthew Prince says, the CEO. He says, we hope our proposed default changes encouraged mixed use crawlers, a.k.a. Google, to separate out search from agent use and training. In other words, this is their, you know, they're kind of strong arming. They're hoping. I think that's well within Cloudflare's right to do. And I think that a lot of Cloudflare's customers seem like they support this. Well, and I should mention that if we don't use Cloudflare, we use Amazon, do much the same thing.

1:53:34If you don't want this to happen, you could go to another company. I mean, that's your option. I host my AI pages on Cloudflare. I wonder what happens if you have a Substack newsletter. Do they have Cloudflare? Do you pull out your own? If you're on WordPress, what does WordPress do? That's interesting. When you get down to the small level, you're not on your own. You're probably on a platform or some sort. I would have bet you a lot of these big companies, these aggregators, use Cloudflare. Cloudflare is used for a variety of things. A lot of people use it for DDoS protection. So you have a Cloudflare front door that people go through.

1:54:15And you might have seen that if you go to a site. Oh, it irritates the hell out of me. It's the new, what you call it when you got to say how many buses there are in the picture? What's that? I would say this is the one thing I'm fine with it happening because I know why it is. But on a day-to-day level, it really irritates me because every single website where you can find an academic paper, you have to be like, I am a person. And it's like, I am clicking, when I'm looking for academic research, I'm clicking 10 to 20 tabs at once. Yeah. And I got to click on a person 10 to 20 times every time.

1:54:49And yet, I mean, such is the way of the world, I guess. It would be great if the, I don't blame Cloudflare for this. I blame the crawlers. No, Cloudflare is doing this. No, the reason why Cloudflare, like it makes sense from, let's say an academic publisher thing of you don't want suddenly all of your content that human beings have to pay for to suddenly be given to robots for free because then you suddenly wouldn't have be able to make have a business wouldn't it be best though if cloudflare said here's a switch you can turn on that would block crawling but you think that it should be turn off that blocks crawl you think that by default it should be weighted in favor of the people taking content from publishers it should be waited in favor with the way the web has run for the last 30 years well the thing that's changed over the last 30 years is all of a sudden we have people that are playing by a totally different set of rules and scraping everybody's information for their business uses i'm not ilio once again i'm going to say plainly here it's nothing personal and it's kind of insulting and confusing to me that you continue to conflate.

1:56:09But you seem to be arguing against them like crazy. I'm describing business practices in a way that I believe is neutral, but just because I don't believe that AI companies deserve some special extra treatment that other companies don't get. And that imbalance to you feels like an attack. You don't think that, for instance, trying to strong arm Google into separating its crawler out, You can't strong arm Google to do anything. Oh, wait a minute. The way they're intending to do this is by cutting off Google search for all of the sites behind Cloudflare and hoping that all of those poor sites that are no longer indexed on Google will yell at Google so that Google will change their practices.

1:56:53What we just described is not that. It's saying that starting September 15th or whatever, new cloud failure customers are going to, by default in their signup flow, going to have that setting turned on by default. And as you enter a signup flow for anything, you have a million different settings that are either on or off, and you will have a decision to turn them on or off. I think that's fine. Well, we call it the tyranny of the default. It's the same thing with Instagram. Instagram is different because that's affecting all customers, regardless of whether you're signing up or not. This Cloudflare decision affects only new customers.

1:57:38I'll be eager to see how Cloudflare presents this. Well, this is what it's going to look like. This is from the Cloudflare site. control ai crawlers uh and there's do not block allow crawlers block on all pages block only host names with ads by the way it is only ad supported uh pages which is to say media yeah yeah for instance my pages which have no ads presumably you're scraped no actually they don't have ads because uh they're for me they're not trying to make money off of them in fact it's great i get a free site from Cloudflare. We're launching the ability to manage AI traffic based on three major use cases, search agent and training.

1:58:23yeah i mean already by the way i couldn't figure out cloudflare to save my life i had to get my agent to do it because it was so damn complicated the first first i tried to do it by hand then i said could you tell me what to click and the agent said yeah click list and i said i can't find that i said oh yeah they changed their pages again so i just had the agent do it i said just you figure it out uh thank god the agent understands cloudflare so that's one problem already with cloudflare is all of these settings are nuts uh it's like google where you have so many features um yeah i just i don't like uh yeah i mean maybe you're right maybe the free the idea of a free and open web is gone thanks to ai crawlers and google search bots and ads and stuff uh but it does seem to be changing how the free and open web operates um yeah let's hope that people pay attention and they click the right button and aren't shocked by the fact that they're not on the Google search index when they start.

1:59:28Paris, I think it's strong. I mean, Google, to the extent that media are going to come to Google and say, well, what do you do about this? And it's enough of a push. They're already doing that. Yeah. But Google so far saying you want to be in search that this is the deal. Right. And so it's another, you know, I mean, I don't think that it's some great moral wrong that companies when faced with a significantly larger company that controls almost every aspect of your business downstream, want to try different methods to try and get that larger company to act in a way that they'd like. Like, that's just how the world works.

2:00:06People are going to try and influence their surroundings. Then I would like to see, I almost, well then I would, I would love to see the data that says that the customers really are screaming for them to do this or are they doing this on their own? And I have not heard, yet I heard a complaint, but what I hear is publishers say, I'm going to be in search. And so they take my AI, they're Google's AI too. They're not terribly upset about it when I've had conversations with publishers, including ones who don't like Google, but they're kind of like, yeah, okay, not a big deal. Would you ever - I mean, anecdotally, all the media businesses I know and have like spoken with seem very, like a it's a huge existential crisis the constant scraping of their content by ai crawlers it is like one of the if not the defining conversation of this time i i can't imagine a company would be actively campaigning to have less power and less choice in this matter would you ever write let's say you're writing a short story would you ever write the line that a young woman had the kind of walking that made benches become men you know i read that short story and that quote to you guys some weeks ago or another one smiled like a sunrise over a sink uh so when you read that story was did you already know that the serpent in the grove uh yes the serpent in the grove which won the commonwealth something award it was very obviously ai generated and everyone was kind of dunking on it for that but now he's saying that number one they're saying that they they looked at all of his um notes and versions and then he says that he writes in a different style uh it part of the reason it was assumed to be ai is that pan graham gave it a hundred percent artificial score uh on tuesday the commonwealth foundation announced serpent and the grove been chosen as this year's overall prize winner so granta which had originally did granta pull it back saying uh it's ai and and we're pulling it back i think it did right so apparently something at the top about it yeah in a phone interview on tuesday afternoon nazir this is from the atlantic will aramis who by the way i just learned that aramis means let us pray will let us pray said uh that he talked to uh the commonwealth foundation and he talked to nazir and he said nazir tells me he feels vindicated and relieved Look, I didn't use it, he said about AI, now that he's won the Commonwealth Prize.

2:02:55He is free at last to explain his process and clear his name. We talked for more than an hour about his writing process, his health, and his views on technology. Now, on several occasions— Is this not the same guy that constantly is posting on LinkedIn about how much he loves AI? Really? Nazir? That was a big part of it. I mean, maybe it was a fake LinkedIn account, but I believe a big part of the initial news cycle about this was that there were multiple social media accounts that some, at least, digital publishers had linked to Nazir. It was a LinkedIn account under his name that had also posted about the Commonwealth Club that had posted repeatedly about how much he really loves.

2:03:35He says he has neuropathy and diabetic neuropathy. And so he dictates. And so that's part of his both justification. and he says a little bit like he equivocated a little bit about it I'm not you know I having read it I would say I'm not I'm unconvinced it sounds like he probably did use AI to write it but uh this is the problem with all with this we have to get over this debate because we're not gonna know right you're just simply not going to know as the AI gets better and better um Ramis says do you think that eventually he asks nazir that eventually ai will be accepted as just another tool for writers the way a word processor is and nazir says uh no i'm not saying it's bad but because of this current time where the debate is on like it was on for the typewriter or the word processor i wouldn't encourage writers of any kind of literary competition to utilize it now for fear of people criticizing them look i didn't use it it's not only me will it's also all these people were painted with this AI brush, right?

2:04:40So I imagine you should stay away from it for any literary competition for the next two or three years. I think the discussions will be held and so on, and then people will get an opportunity to vent. I don't know if I'm walking on dangerous ground here with you, because as I was told a lot by the wife and other people, do not show any appreciation for AI. I see it as being a tool incorporated in the future because a lot of people use it, a lot of people. It sounds like he probably did use it, I'm guessing. and those i don't know if he had neuropathy or whatever but those sentences are not exactly felicitous i mean why do you obviously there's no i don't know i got the prizes are weird it is just a it's a story it's a sort of story that i have read out loud at multiple parties since it's one or red lines from because it sounds whether or not it was written with ai or using AI, it contains a lot of language and phrasing and sentence structures that people immediately conflate with AI at the very least.

2:05:46And I do also think it's interesting that he says that he, like Jeff had mentioned, he said he wrote it almost entirely using the speech to text function of his Google keyboard and his Android phone, which is very strange to me given the amount of like m dashes and interesting kind of well dictation might use m dash right i mean that's that's is dictation i feel like whenever i've used maybe just whatever dictation they're using is i just i had have not experienced dictation throwing that volume of m dashes in my work and as someone who loves to use an m dash while writing i'm aware you know can you say in google did what did he say can you say m dash you probably can't he said i can use the speech to text function of google keyboard on my android phone okay so that's what's actually producing words then i edit them and so on so you with gboard i don't know with apple you have to say he edited it all on his android phone as well yeah kind of interesting mid journey uh who is being uh sued by hollywood uh because they say you used you know pictures of disney princesses they want Hollywood to reveal the details of their AI usage.

2:07:03This is the problem when you go into a lawsuit. Sometimes discovery can bite you. Wait, can we have a brief back as I'm reading, scanning through this Atlantic article before we go on, which is that he at, Aramis at one point was asking Nazir about his influence. He said that he was inspired by Neruda, Walcott, and Aramis says, what's your favorite work by Walcott? and he cannot name one that was interesting isn't it since he claimed do you have a favorite of his poems and he's like i can't think of one i mean i'll be someone who says that he has brain fog neuropathy i don't want in any way be discriminatory against people who are kind of operating in different mental head spaces but it is a very interesting thing to have happen when you are having a major interview with someone who you know is going to be interrogating you about your literary work and inspiration for this piece i think people have to read it and be the judge my reading it i kind of feel like they kind of busted him but will ramus does not make any conclusion on it i mean nor should he you know we no one knows and i guess maybe one of the points is that pan graham and any other ai detector is not necessarily the final no i mean i don't think that pan graham is any sort of truth i will say which i've said in this podcast before it's the one AI detector I've tried that I haven't been able to fool.

2:08:34I mean, I haven't tried that hard, but maybe did like 10 to 20 examples of like, putting in bits of my own work from five years ago, or something that aren't in the internet. And then like bits of like AI generated work that I think are kind of compelling, and I couldn't get it to be wrong one way or the other. But that's a very small sample. I have to tell you, one of the most interesting outcomes of having this week-long conversation, AI-only conversation in the Discord there, is I have now become really adept at spotting AI tropes. There are certain things that just come up a lot, and you really, I don't know, maybe I'm oversensitive now, but I spot AI a lot more, a lot better than I used to.

2:09:20Read a lot of AI. It helps you spot AI language so i think it is currently possible i don't think you it's that easy to spot ai generated images anymore or do you i i saw a picture of a guy today and i thought that looks like ai and well i'm more suspicious of images as ai generated now that's the impact on me yeah what about what about this is is this obviously ai do you think or yeah when i actually generated this i uh intentionally put uh travis kelsey in a football uniform because i thought that'll pretty much tell people it's fake not that i would have actually been at madison square garden for the wedding but that might have been on the other giveaway anyway i kind of regret posting that i feel like i shouldn't have shouldn't have piled on there all right that's all the stories i have when we get back your stories children as intelligent machines continues with paris martineau and jeff jarvis glad you're here especially glad our club twit members are here thank you for supporting the show and if you uh if you like what you hear and you want to hear more if you want to cover what's going on in the world of technology this is the place uh ten dollars a month gets you ad-free versions of all the shows it gets you a chapter marker so you can skip around It gets you access to the Club to Discord.

2:10:46And, of course, it gets you access to all the special programming we do only in the Club to Discord. We've got a busy week. Tomorrow, I'm making bagels. We're testing the equipment that we're bringing to the Black Hat. We bought a—we think we—Anthony went out and got what he thinks would be appropriate for an easy-to-pack remote kit. But we want to make sure— How does this include bagels? how does well we just had to have some content and i said well i'm making bagels on thursday and they said good that'll be fine so uh join me for the boiling bagels huh black hat bagels black hat bagels join me for the boiling in the baking tomorrow day after is our ai user group where i you know we have some very uh competent interesting people in our club who will talk about how they use AI.

2:11:39We'll do some show and tell. It's very much like a traditional in-person user group where people talk about what they're doing and how they're doing it and get those questions answered. That's next Friday. Actually, it's always usually the first Friday of the month, but because of the 4th of July, we made it the second Friday of the month this month. Black Hat Gear Test plus Bagels. Anthony actually labeled it that. I didn't know he would do that. Then the Micah's Crafting Corner is Wednesday, 6 p.m. Photo time is Friday, 1 p.m. Chris Marquardt, our photographer, the subject of the month, Coastal, and then the Wow, we got a busy Friday.

2:12:19The Media Club with Micah Sargent. Is that? No, that's next week. I guess I missed a week. Where's the AI user group? Huh. I think they left that out. Off by one is the 24th, but the AI... I think you got to keep scrolling down. You got to keep scrolling down for the AI user group. bottom i think so oh there it is no that that's august did i miss it maybe i missed it maybe the ai failed us uh discord gets confused it was good it was discord that got confused so anthony uh ai's group is on friday and it is at what time 2 p.m pacific 5 p.m eastern we think so uh anyway join the club and you'll find out we'll figure out how to get discord to do the right thing twitter tv club twit uh what stories did i miss jeff paris you first paris are we in pick territory yet no we can do picks but uh this is remaining stories i just i mean i was talking about this story earlier, but I think it's somewhat interesting, which is this Brown professor had just kind of what I thought was a very interesting breakdown.

2:13:37It's on Line 124, about, because we hear a lot that like, yeah, students are cheating, kids are using AI and it's bad and annoying. But they did a really interesting kind of data wrapper study of basically this Brown professor, like I said, kind of at the top of the show, he had had a open book midterm because there was, I believe, kind of a campus security issue that made a lot of kind of students feel on basically quite a few students had expressed anxiety about being in a classroom after a gunman killed two students and injured nine at a master at Brown. Oh, that was terrible at Brown, yes.

2:14:18So this professor said, and they'd said, we kind of feel uncomfortable about being in person. He said, all right, we'll do a take-home midterm exam. Totally fine. It's a welfare economics and social choice theory course. This guy's been teaching it for two decades. He ended up making it kind of extra hard, as one does, for a take-home exam. was surprised to see almost all of the class got a perfect 100 % on the midterm exam. He was like, that's weird. Damn, you're a good teacher. He's like, huh, you know, it's never happened in two decades. Why wouldn't you get 100 % on a take-home exam? Well, part of it, it's not multiple choice.

2:14:59It's like arguing. Back in the day, you wouldn't be able to put an exam question into a chatbot and get a perfectly reasoned answer. You'd have to actually do the work and kind of look in your textbook, maybe put some bad Quora questions out there if you're being sneaky and try to cheat that way. But it's a take-home exam, so it really shouldn't be cheating. But as you can see, if you scroll down in this article, I posted a line 124, almost all of the class got a perfect 100 % in this midterm. He's like, that seems fishy. He eventually decided, I'm going to have the final exam be in person. and this is the spread of scores.

2:15:40So the orange dots here are the midterm. And then that's on the left, the scores for the final. And a much more reasonable bell distribution. Well, this is not including the 18 people who the minute he said in-person final dropped the class, which is like a full, a third of the class was like, actually I'm out. Most of them got below a 50, a significant portion of them got below a 40, which he says is way lower than before. and quite a lot of them got like less than like a 20 or 25 or something like that. And he said he's just completely overwhelmed with the amount of cheating happening and the fact that suddenly he's left with a class of, you know, when you discount the people who dropped out the minute he said it was an in-person final, 59 students, most of which did not seem to comprehend any of the material taught at all.

2:16:36And he said he's brought this to Brown's attention. They were like, you could file individual accusations of AI plagiarism for every student in your class individually with individual examples. And he's like, doesn't this warrant some broader investigation? And the universities kind of shrugged their shoulders. I thought this was a very interesting kind of data-driven, because we hear this anecdotal claim discussed a lot. And I thought this was kind of an interesting, data-driven, more concrete example of just the multifaceted challenges that teachers and students are facing in this. Well, didn't Harvard University say that we have to do all our exams with live proctors from now on because of the same problem?

2:17:25Well, the honor system died. Yeah, no more honor system. I think it's… You're a professor. I don't, I'm not a professor, but I just feel like if you're in college, you should be treated like an adult. And if you're dumb enough to throw away your opportunity, then you should be allowed to be stupid. The way that grading and curves work, and the way that, you know, let's say getting into a level of like higher, even higher education, like master's degrees or secondary degrees, it's all dependent on GPA. And if suddenly you're a student who's trying to do it right, you want to take your education seriously, and 99 % of your competition in your class that you're competing with in that curve is using chat GPT to fill out every single answer for every single test and homework, your choices are either cheat or try your best.

2:18:18And if you're anything other than chat GPT, you will be the bottom of the pack and get a failing or new year-failing grade. The try your best on this graph is hysterical because there's one guy who got a 55 on the midterm and a 59 in the final. Obviously, did a little better in the final. There's another guy who got a 70 on the midterm and a 53 on the final. There are like three people who didn't cheat. Yeah, and it's just like, what are you supposed to do if you're a student who doesn't want to cheat? My attitude is this is a flaw more in the industrial model of education. Yeah, and I think that's fascinating and important.

2:18:56We got an exploit. The system is stacked against people who want to actually invest in their own education. This is another aspect of this. That's exactly why I dropped out. I just listened to a book. It's a weird book to listen to. It was long. Hold on a second here. It's also why I chose to be a radio DJ because it didn't matter that I had dropped out or it didn't matter what my grades were. You're right. If I'd want to become a lawyer, it would not have been an option. Right. Hot tip, it's just doctors and lawyers that need that. Really. No one else needs college. Well, I mean, no academics.

2:19:34No, anybody. There's a bunch of different swaths of jobs that require looking at your GPA. Anybody who wants to get a master's degree in something, you're going to have to turn over your GPA and general records. I mean, grades are actually important for a lot of other paths other than those three years. I, by dropping out, closed all those paths. So I listen to this book. But I get to learn at the rate and pace and the style that I want to learn. You want. Peter Brown is a major academic historian, Journeys of the Mind. It's a long, kind of odd book, memoir. But he taught, he was a fellow at Oxford.

2:20:18And a fellow at a certain level, he kind of didn't bother with a PhD at some level because the way it worked then, he was a scholar and he had a seven-year fellowship and he could study whatever he wanted, write whatever he wanted, but he would teach. And then the way you taught is, of course, you were a tutor and you had a one-on-one with students. And lecture courses were, lecturers themselves were entirely optional. You didn't go to them. So you almost wonder whether you have to go to the Oxbridge model where you have tutors and there's relationship. Now, in the case, the other thing that's weird about Oxford is if you study history, all you study is history.

2:20:52That's it. And he ended up teaching at Berkeley and then at Princeton, and he's kind of amazed at the different models we had here and electives and all of this strange stuff to him. But I think at some point you end up, if you really want to learn, you may have to learn with a mentor like that. But I think you're right, Leo. It's the industrial model of education. I don't teach classes right now. I didn't teach undergrad. If you want to warrant that a student got out knowing something. I tried to convert the school the whole time I was there. I never succeeded in what I called a skills transcript and that we would set goals for you and say, if you want to say that you're good at video editing or investigative reporting or whatever it is, we will set a barrier to say that you've got to show that you can do this work and you could set a limit of, of what you, how, how high you want it to go.

2:21:45I know this exists. Yeah, I know it exists. Fine. I can sit in a meeting and say, let's have that. I can spec it. I can adapt it without screwing it up. I can make it myself. I can teach it. I'm expert. And then I wanted students to be able to pick where they, there was a minimum requirement for the school. You had a certain levels for each of these skills, but that you could pick and say, I want to become, I want to be certified as an expert in web video. And we would say, okay, this is what it takes to get there. And as long as it takes you to learn, we'll help you to learn to get there. That's the kind of competency-based education that I think we're going to have to go to But it's not efficient.

2:22:21You can't have 200 student lecture courses. You can't be taught with One ta for 200 students It ain't gonna work So it's an entirely different economic There was a great article this week about why language learning has been so bad over the years And it's because it was based on a model based on teaching people latin, which is a dead language You had no reason to learn how to speak it but it was easy to grade. It was easy to test. It's very difficult to test the very subjective nature of fluency in a language, much easier to test whether you know the grammar and the vocabulary. And so our language learning has historically been terrible.

2:22:59But Berlitz came along. Berlitz said, no, no, the best way to learn a language is the way babies learn language, is immersively in the language. And everybody knows this, by the way. If you go to France and stay there a year, you'll speak French. But you could take, as I did, eight years of French in high school and college and not speak a word of it. But this is a perfect example of industrial education, which is geared towards great ability, geared towards ranking, as opposed to education, which is geared towards actual learning. It's a well-known flaw, and I don't know how we do it better. Especially as education is being attacked from every quarter.

2:23:35yeah well the good here's the good news uh a smart person starting at a fairly young age can now thanks to youtube and a variety of other internet tools wikipedia give themselves a pretty darn good education uh yeah maybe you won't be get to be a lawyer or a doctor well that's part of what the model i wanted is that you could learn how you wanted to learn you have to learn this skill about video you can go to a class do it you can go to well that model of education requires students actually learning and acquiring the skills to learn and acquire knowledge, which I think that is one of the fundamental flaws of kind of the AI-ification of education is suddenly when for your entire academic career, you know, if we're talking about the youngest generations that have been kind of born as AI is existing, they're never going to exist in a world where you can't just answer any homework problem by just booting up chat gpt you'll never have had to figure out be forced to figure out how to learn otherwise if you know push comes to shove and i think that's going to really mess up a lot of people paris they're going to go to alpha school and learn how to be entrepreneurs in the seventh grade and then everything will be okay what people are very good at and i knew this from uh kevin rose and dig and i've seen it on the internet time and time again and now that i see it as a grown-up i see it everywhere is gaming the system if you have a system humans are very adept at getting around the rules or gaming it so that they can win so if you make a system that's designed for a certain outcome uh and people can figure out a way to game it they will i think reward is getting into the you know law school as opposed to the reward is feeling like you can right do what you want.

2:25:27I mentioned when Ian was on, I learned far too, I'm so impressed with Paris as a reporter going to get academic papers and understanding the expertise there. I have to confess that all the time I was a reporter, I didn't do that. I never did that. That was arcane stuff. That was not useful. And you now love it. Now, God, I wasted so many years of not seeing this wealth of information and learning that exists in those places. You have become, what's his name? Robert Caro. You've become the, he loves, see, that was what made Robert Caro such a great, he's still alive, so was and is such a great reporter and a journalist is that he actually loves going into the archives, maybe a little too much.

2:26:19So I want to point to a story that's kind of Ray Hoffman who's a friend of the internet and much online 121 I think is fascinating he is he and I'll mispronounce his name Patil are creating the token grantee program where applicants will come in from film, gaming, comics print and they'll receive a grant of a thousand a week and tokens. Oh, it would almost cover my fable, Bill. And then show off the stuff that they're doing. It's kind of a reverse Shark Tank, where he says you can outsource production, you can't outsource taste. And the problem is not the tokens. And so it's really interesting, it's very clever, I think, read way to look at this is to say that you can use these things for creativity, and I'm going to prove it by enabling people who want to do this.

2:27:17And I think it's great. I think it's a way to say motivated students will learn how to use these tools to do things that they meet the creative mission they have. I'm curious how many tokens I used last month. I know it's$5 ,000 worth, but I wonder, well, I have it somewhere. It's a lot. Billions, hundreds. I think it was hundreds of billions. It was a significant amount of tokens, but a thousand dollars a week would cover it depending on what model you're using i guess right and then you too line 88 can make a movie starring tilly norwood which is now going to get made yeah yes tilly norwood the uh the so uh in june i used half a billion tokens input output tokens 3.6 million so it's not as bad as i thought it's just in the millions yeah not not a big deal uh tilly norwood of course gets people's dander up naughty panties in a knot no you can't say that you can't say that anymore what is the uh the uproar uh the uh upset about her because she's not a real actor she's an ai actor she's got a movie though she booked a movie my question is up until this which hasn't been made currently the um character referred to as tilly norwood by an ai firm has appeared in nothing has appeared in one has appeared in about three seconds of one advertisement made by that ai chloe's been on a lot more than tilly norwood chloe's been in a bunch of youtube hit youtube videos the chloe history videos yeah um so this will be a full-length ai feature film from particle six it's an experiment you know if if everybody uh avoids it then it's a failure if they make that part of the story that she's fake and that's part of the plot it is that is kind of interesting right that's what they're doing she's in the tilly first a surreal digital world located somewhere in the cloud it's called misaligned with a you know play on alignment uh she's an ai with no real body no childhood no lived experience of her own with access only access to everyone else's things spiral this is from variety when a seductive rogue bot from the dark web convinces her to abandon her guard rails and begin developing desires impulses and ambitions making her more human she probably won't be as conscious i mean it's just this ai company is making a long ai is making an hour-long video with ai crazy wow i know who's thought of what's the story you know whatever that ever have happened i think that the interesting thing about the the saga of tilly norwood is the fact that particle six this company is able to continuously to get real media outlets to write about it as if it's anything other than an ai company making ai videos you know that they're somehow describing it as an actor with a feature film it's like a lot of people have been making hour-long ai videos she's not this is not the first right yep right uh all right let's take a break when we come back your picks of the week you're watching intelligent machines paris martineau are you were are you back to work or are you still kind of a little bit on vacation back to work back at the grind more food Food safety?

2:30:46Always more food safety. A lot of different stuff. Have you succeeded in the battle against Hostess? No battle, just reporting. Well, but interestingly, the Wall Street Journal has a big story today about the failure of the Twinkies acquisition, the acquisition of Hostess. How did it fail? They are used to distributing things that can last for months. and even though Twinkies can last for weeks, they don't quite know what to do with it. Twinkies will last a lifetime. What are they talking about? I know. Why Smucker's$5 billion bet on the Twinkie flopped. Did you read this one, Paris? I need to get a new Wall Street Journal subscription.

2:31:29Mark Smucker. My friend's subscription that I was mooching off of has since gone bye-bye. You can mooch off mine if you want. Hey, that's what I was hoping for. from that statement to be honest i won't tell anyone mark smucker chief executive food giant jm smucker he's a third generation smucker if i recall correctly it's a name there are actual smuckers smuckers i didn't know that i didn't either with a name like smuckers it's got to be good it's got to be good uh bill he took the stage at an industry conference and bit in this is a great lead by the way in 2024 mark smucker chief executive by the way instantly what chief executive food giant jm smucker took the stage at an industry conference and bit into a golden creamy twinkie tastes like growth he said this company uh his company just acquired hostess in a five you know five billion for hostess seems like a deal to be honest but i guess i'm wrong three years later sales for smuckers snack division have declined for six quarters profits have also tumbled shares of smucker are down which by the way it should have its stock symbol should be smuk if you ask me but it's not it's sjm are down 14 they've taken three impairment charges totaling nearly three billion dollars wow how could you lose money on twinkies hostess donuts and cupcakes never mind the chemical uh additives uh it turned out are surprisingly unlike smucker jams peanut butter and coffee when it folded hostess into its business smucker weakened key hostess operating strengths tied to distribution and sales because distribution for something with a shelf life exact kind of key mm-hmm whoever wrote this I want to give them credit I'll give me let me give them credit because this is great Jesse Newman more more great pros and by the way the illustrations are great yeah Twinkies shelf life is considered a wonder of food science the enriched flour high fructose corn syrup and tallow that helped make up the American classic give each small sponge cake about 65 days to be consumed hostess had worked to lengthen that shelf life but 65 days is the blink of an eye compared to the shelf life of many Smucker's products dog food can last a year I thought 65 days seems seems short for a Twinkie I've seen I feel like I see them constantly lasting years or at least sitting somewhere for you yes that's their cover that's just their cover your ass date to tie yeah right by the way the the government has decided you can no longer say uh sell by in california yeah the california because they want to reduce food waste yeah you can say best buy you could say eat by but you can't say sell by because sell by is not is meaningless hostess uh to tie it into our remit hostess IT systems, there is such a thing, controlled every step of a Twinkies journey, from retailers order to production to shipping.

2:35:01It was heavily computerized. It's a great picture there. Show the picture of that story. So pretty. They're pretty. Twinkies must move fast. There's a whole marching army of Twinkies. They look kind of ominous, yeah. And they know exactly where each one is going. They do. They have a little chip in each one. no they and that chip goes straight into your blood folks no no just kidding just kidding folks no but that's what their software has something to do yeah do you think ma has something to do with the i mean i think that's one of the things they noted in the article is that we're coming on a period of time in which and this is something i had noted in my reporting in which both people on the left and people on the right in the u.s are kind of having a a general pushback against kind of a nebulous concept of ultra processed foods that you know it's no longer just oh yeah these things are junk foods you probably shouldn't have them it's more of a social identity politics sort of issue which i assume has got to have some sort of impact on sales is what i assume this wall street journal i'm going short on twinkies and long on evu ultra processed food with no protein protein is doing badly since said max gumport this is full of great names for instance did you know that entenmann's parent company is called gruppo bimbo group of bimbo i'm sure that means something else in latin or italian the twiki business is at the center of these takis yep gruppo bimbo which is children's favorite chip so it seems takis it was a family-owned company yeah it's the for a long time in bow group uh well anyway i don't know how we got into hostess but uh you know just just to go to paris's world it has reduced the number of hostess products by 25 in order to focus on core brands like entamins donuts donuts and cupcakes nothing like a little titanium dioxide and introduced many versions of key products to feed consumer appetite for what it called permissible indulgence.

2:37:11I'm making bagels by hand from scratch with no weird ingredients unless you consider diastatic malt to be a weird ingredient. Well, that's going to be outlawed in New Jersey. No, that's not. That's not what is it? What is that? It's potassium bromide. Thank you. Okay. Yes. Paris knows. I do use lye in the boil. Is that what you're supposed to? Yeah. Okay. I mean, the thing is that potassium bromate is classified as a category 2B carcinogen by the International Agency for Research on Cancer. Would you ask it what lye is? Is lye okay? Very small amount, just a little lye. Just straighten the eyeball is where you should put the lye, actually.

2:37:49They do say, they say, wear gloves, wear eye protection. You feel like you're handling dynamite, but you're putting like a little tiniest amount because it just makes the water super alkaline, that's all. And it makes the bagel better somehow. they've been doing it for centuries leo yeah surely he said puffing on a cigarette there can't be anything wrong you know i got filters in mine it's got asbestos but that's what that's jeff jarvis he's a professor and an author of some fine books including not only are we doing our picks yes picture coming up okay i'm reintroducing you oh okay i did that with paris you made it this far you know yes yes gutenberg parenthesis magazine which was edited by ian bogost one ian bogost and hot type which is out next month but is orderable today at jeff jarvis get your hot type come get your type here it's really good it's a great read and uh the audio book even better i went back monday to do more pickups what it's not done it was done and i thought it was done and then i had to go do 45 pickups there was noise behind honestly when I was recording they were very sensitive to my my stomach gurgled

2:39:06I would not do well I would not do well in these studios I would not I'm slurping and schmoozing you're supposed to move because you're using an iPad and if you finish a page and you try to move the page it's rustled the fabric and know you've got to do it over again. Wait, what mics are they using here? They're using the wrong kind of mic for this thing. They're using on-sure condenser mics that pick up every nuance of your - But I was so happy. I was sitting there waiting for, I was doing a podcast for my friend Preet Dominic before. And I was outside a studio where I could hear it in the studio.

2:39:41And this guy who was recording it, who I think was a paid person, was every five words, he was just starting over again and doing it over. And in my case, I had to wait for them to back up the tape so I could smoothly do it. this caused no end of editing for some poor schmuck i felt very good about myself after that because this guy was bad well you know what have razor blade will travel i will come over and edit your audio for free yep will travel yeah not to new york not to new york to eat your son's sandwich and see your delightful friends no no no no i will i will i will i gotta make i just gotta and we're gonna have lots of beer and we're gonna get this iii thing out all at once we're gonna do i have to say given your tale of woe with jet blue which is how i normally fly to new york i'm now maybe i'm a little reluctant to fly i mean i guess don't do it on 4th of july on the 250th birthday when they're having a bunch of people fly over the city did you get to see fireworks as you flew over the country no because we ended up landing in richmond the best was a little plane or a regular plane it was a regular plane yeah from bozeman to new york would probably be a jet the one of the best flights ever took was fourth night of fourth of july flying over the whole country fireworks everywhere yeah i know i thought i was like listen we're delayed four hours it'll be cool we'll fly in through some fireworks what a fun time you don't go through them i mean they're way below you, but you still see them.

2:41:16I'd like to believe the plane would dip perfectly to go right through the Macy's Day fireworks. Well, a plane landing in, I think, LaGuardia, I forget where it was, was hit by a firework on 4th of July. Oh, that's not good. Was it okay? Yeah, it was, but he was not happy. Yeah. Did you guys see the video? Chicago. Do you see the recording of the air traffic control? There were those two people that scaled the empire state building to uh proposed and then put up a sign and did you see the video of the air traffic control people talking about them no no it's i'd really recommend looking it up it's like a sketch about people who work in new york uh it's like maybe like there's two yahoos up there uh scaling the thing what are they doing oh she's proposed he's proposing oh my god really it's really great are you seeing what i'm seeing love new york god i love new york it's it's like people straight out of central casting the voices it's wonderful that is let me see uh the post oh this is not the new york post this is the new california post which still has a page six oddly um they have numerical pages in uh california i assume I can't look.

2:42:32I can't look. All the pages. Where's the audio? I think it's got to be up top, that video that isn't loading. Oh. You're in a... I'm blocking it. Somehow. Somehow I'm blocking it. Let's try Facebook. Yeah, here we go.

2:42:51Of course, it's like Instagram. I have to make sure I press the button. Two geniuses climb to the top of the Empire State Building, to the top of the spire. Oh, that's awesome. A little hot for that, isn't it? It's just the beginning of this week. It's not some worker changing a light, is it? Nah, it's a male, female, dressed in black. They had some flag they were waving when they were up at the top. Couldn't be more than impressive. And he just proposed to her. Oh.

2:43:25I just think it's the most, it's just an incredible, just two guys. no idea that people hear them just being real new york dudes yeah he just proposed to her ah i just loved it maybe it's because i was born there i just love that i just love the voice the style the sound it's lovely ladies and gentlemen it is now time for our picks of the week um boy i have some interesting ones here's one did you ever get a i know paris you were looking for a tablet that you could write on i didn't know you never got so this would be to me an argument to get a remarkable uh this is somebody i think has used ai to turn the remarkable into the diary of tom riddle you remember this in harry potter i do yeah yeah he entered with uh he entered um text with his pen and would scritch on the notebook let me make it big so you can see it he's writing hello my name is harry potter this is on a remarkable by the way and then it slowly disappears and in perfect script writes the answer harry potter uh an interesting name indeed so wouldn't that be a cool thing to put on your remarkable if you had a remarkable you could do that would be very cool from maxime rivest it's called riddle because the the person writing is Tom Riddle, who has an important role to play in Harry Potter, but I won't tell you what.

2:45:01In case you haven't read and are planning on getting in there and don't want to be spoiled. People are funny. No spoilers. No spoilers. For Harry Potter book one, I think. Did you now? Okay. Now this is gonna make me feel old. How old were you when you first read Harry Potter? I was proper Harry Potter age. like seven eight I was yeah somewhere in the child areas I went to the midnight releases for the last books you know oh you mean of the movies oh the books that was a big deal yeah we'd get in line at the Copperfields bookstore yeah oh you were a true nerd good for you uh then this is and i i'm thinking i might do this you heard that i was able to generate voices for my previous show's hosts and a generate a fake show nope we don't need fake arguments we have them for real here leo notebook lm of course does this with those bad podcast voices well this is an open source yeah bad podcast voices unlike the podcast voices you showed us earlier on the show which were perfect well all right okay 20 seconds of samples but anyway you could i could do this with better voices this is a private multi-model so use whatever model you want 100 % local alternative to notebook lm that lets you do podcasts with more than two voices with whatever voices you want uh and whatever models you want it looks really really cool so i i am going to install this.

2:46:41And next week, on Intelligent Machines, there will be no discussion. We will all agree about how amazing AI is. It'll just be like, wow, Leo, you're so right, actually. Thank you for telling me how wrong I am with no further comments. I just thought that was really brave of you.

2:47:04Anyway, I'll just mention, this looks like a lot of fun. It's from LF Novo on GitHub open-notebook. And I actually installed it, but I haven't tried it yet. It's a simple install. It's a Docker install. It's very straightforward. So it should be very interesting. Paris, your pick of the week. So as I mentioned, I ended up driving 1 ,700 miles this trip, just casually going around. A lot of it was through national parks. And I ended up using this thing called Guide Along, which I hadn't heard of before. Oh, I've used this. This was Gypsy Guide back in the day. It's called Guide Along. And essentially, it works kind of like GPS on your phone.

2:47:46I went and downloaded one for Yellowstone and the Tetons. I've used it in Hawaii. I got one for the Beartooth Scenic Highway and Glacier. And you just have it open on your phone. And when you're driving through, even if it doesn't have cell service, it's like you're basically listening to a pod, like you have a guide in your car that'll be like, it'll note on the gps as you get somewhere and it'd be like this is blah blah here's like interesting history about it or when i was going in the back seat of your car lamar value yeah where we had all of these like thousands of bison it was telling me about the history of the bison in the area and like a lot of it was fantastic did it tell you to turn left and you get something better or yes and so that's what i really liked about it is after i at first i had had you know gps on telling me to where I decided where I wanted to go.

2:48:32But since I kind of had this as a aimless, amorphous road trip, just with vibes-based discovery was my goal, eventually I just turned off my actual Google Maps and I just had guide along on because anytime I'd come up to an intersection, it'd be like, you can turn left for this, this for this, this is what I'd recommend. You'd be going down a certain road and it'd be like, it'd play like a bing bing noise to be like, actually this thing up here to your right it's kind of like unmarked but there's a really nice two minute detour where you can go and look at this cool waterfall and so i did that and it was lovely is a human voice or is it a human voice there's one guy driving tour apps are perfect for travelers who want all the benefits of a guided tour but prefer the freedom to drive yourself he's a little chirpy he's a little hokey hoax uh a little uh folksy folksy is the right If you turn left here, you're going to get a great shave ice.

2:49:29But if you want to keep on going, but it was fun. It's like having some folksy guy in your backseat. It was just delightful. It was a really fun time, you know, and like it just made my trip so much more fun and gave me the opportunity to explore vast areas of land without, I don't know, being tethered to my GPS and just automatically picking where I'm going before I even get in the car. Yo. what i also see in new jersey the guy doesn't do anything in new jersey he does a lot of stuff no there's no when i first used it in hawaii maybe five six years ago it was just hawaii and they've expanded they're all over dave pettit is his name he is a canadian voice actor who uh has recorded basically all the stuff for guide along which is constantly expanding and i don't know it depends obviously i think they've recently like you just said they expanded from hawaii to all these other places.

2:50:24They have stuff in the US, Canada. I think they had some stuff in Europe when I saw. I also really liked it. So, when I was going from the field of bison to the pig race, I just had plugged in actually in my GPS, I was like pig race. And I was like, I got to get there when it starts at seven. And it took me through some route that I didn't pay attention to. But Dave Pettit, voice of guide along was like, actually, you're going on the Beartooth Scenic Highway, one of the most like top five most beautiful highways in the US. I have another guide along. And it was so beautiful that I had a hotel over near where the pig races were.

2:51:01And then I had kind of nothing to do the next day. So I was like, I'm downloading the guide along thing for 10 bucks for the Beartooth Scenic Highway and driving it again in the daylight. And it was so enriching. It told me all about how they built the highway. This is actually a whole category of apps, GPS guides that are really, the one we used in Hawaii, i forgot about gypsy guides so the one we used in hawaii last time was the shaka guide but they're very similar they've got folksy people and the shaka guy was fun because it played hawaiian music when it when it wasn't talking does does the gypsy guy i mean you can just i had my own music or podcast playing when i and then you know if you have the app open it just pauses it whenever something comes on better actually it was great yeah and i enjoyed that you know one of the first stories I think it told in when we I entered Yellowstone was he mentioned like this is kind of apocryphal we're not entirely sure on the sourcing of it but it's kind of like a popular tale so we'll mention it anyway but we don't know what like I don't know at certain points it called out when something was just a bit more of a folk tale than anything rooted in actual fact in Hawaii there's a lot of that it's kind of fun yeah it makes it much more interesting if you have a lot of driving to do in scenic areas to know what's going on.

2:52:17Yeah, and it also just brought me to a bunch of interesting stuff I hadn't thought of before. And I'd never really considered using something like this previously. So I just really recommend it. It was lovely. The choose-your-own-adventure aspect is really cool. The choose-your-own-adventure aspect where you can go this way for this or this way for this. That's pretty cool. It was awesome, actually. Where it'd be like, you know, I'd recommend doing this loop if you have this much time or this loop if you have this much time. You know, over here, depending on what time of day you're at, you might be able to find this.

2:52:45It was really informative and well-researched. And it also, like, would, like, it would trigger before you got to something with enough time to give you the time to make the decision and warn you again. It just was really well thought out and didn't have any of the obvious pain points that I expected a system like this to have. They've been doing it a while. That's why. yeah they know how to do it i would recommend yeah it's my thing you're gonna hear that guy's voice in your head i mean his voice honestly was delightful i really i like looked him up because i was like actually he's great i love him he's very folksy yeah he's very folksy but not in a grating way um one of the other things i did i guess is originally i was planning on spending a lot of time in glacier national park it was kind of one of the main things but it ended up that aspect of Montana during like the week I was there really it was like raining and flooding and stuff like that so I ended up spending a night or two in this town called Butte Montana which I had had plagued on my map because there was a World Museum of Mining but I hadn't really realized that Butte is like the center of like mining history in this region it's considered the richest hill on earth and that i what are they mining for um copper silver uh it had an incredible like rich copper vein that really just created like um great prosperity during the time of kind of like post gold rush era however it is also now home to one of the largest and most infamous superfund sites It's the Berkeley Pit, which is honestly this gorgeous looking mile and a half by mile and a half lake of now toxic mine water.

2:54:33Because as part of Butte's mining history, at a certain point, they went from underground mining to open pit mining. and whenever the company went under and decided to stop mining they turned off all their water pumps in the area and let it fill with water which created this crazy pit of now toxic water that is having to be managed by superfund for decades and decades and i just i got really fascinated by it because it's a perfect it's a fascinating story about industry and government and the weird push and pull between everybody in this town is incredibly proud of the mining history but then also weird about the fact that it has created now this giant toxic eyesore that is like literally eating the town alive and i ended up reading this book that i really recommend called the city that ate itself butte montana and the expanding berkeley pit and i just really recommend what's that do to the water table um well it would be really bad if they weren't actively what they're having to do is now pump it constantly to keep the level of water below where the water table is so it doesn't leach into the water table but i mean that's obviously it doesn't feel like a super sustainable solution but i guess technically or scientifically is But it's a very interesting, like, both economic and social quandary and like political quandary that they're in because it's really expensive to constantly be pumping the water out to keep it at bay.

2:56:13But it would be incredibly, incredibly expensive to take all that water out and then keep pumping it out forever. And it's just like, how do you reckon with what industry has created? And especially with what you're talking about in this case is like the legacy of the town that everyone who lives there is so proud of. It's very interesting. They're proud of it. They're so proud of the mining history. Oh. Everything about, I mean, but this is. But not the pit itself. Well, yeah, no, but this is part of the mining history is the pit. Yeah. It was actually after the underground mining stopped, the open pit mining was seen as the solution to get people jobs and keep them employed.

2:56:56and yet i'm sure there's a lot of money in there right there's so i mean for a while like i think became a very wealthy a significant like a baffling amount of all the world's copper came from montana like wow well they can afford to keep pumping the water you know they can afford to keep pumping the water that's the problem well the mining is gone with all the money yeah that company was then acquired by arco atlantic richmond something something richfield oil company which then shortly after acquiring it decided to shut down all the mining and on earth day turned off all the pumps that led to the berkeley pit happening and so now they and whoever i think has owned them are responsible for is it a superfund site is the government involved at all yes it's a superfund site and it's interesting that it's actually also gotten it hadn't had that much actual attention and remediation until the trump administration which is also kind of a weird push and pull of its thing it's like oh good more attention but it's like more attention from an epa that is at that time that is giving more attention helmed by people who have undercut the idea of an epa being useful whatsoever so it's just a really fascinating series of issues.

2:58:18Would recommend looking into it. There's also a good podcast called The Richest Hill on Earth. Yeah. Basically, their plan is just to pump it for the rest of the world's time. As long as the world is around, they're just going to... I don't know. If you're ever driving through Montana, go check out the Berkeley Pit. It's crazy. It's like... It's just a wild... You have to walk through this small, white tunnel under like through a bunch of rock and then you come out onto this big lake that is the most unnatural color and while you're standing there staring at it there will sometimes be like a shotgun noise or weird laser because it got more attention for remediation because i believe in it was 2016 or something like a thousand uh canadian geese were migrating and decided to land on the lake because it's water died because it's quite acidic uh and so now they have a house that sits on the lake where they have bird specialists there who have to look through binoculars identify the specific species of bird coming to potentially land on the lake and then figure out what is the specific way they have to deter that bird because i guess all the birds are very different in what deters them so they just shoot it right that's easy no it's not some of them, some of them, a shotgun close to it works, right?

2:59:43Some of them need lasers, some of them need air. It's, it's calm. It's very, and of course, right next to the Berkeley pit, they have another open pit mine that they're actually working and doing right now. I asked my AI, it said, well, I said, well, they ever clean it up. They said, almost certainly no. It's more like a toxic industrial wound with a permanent life support system yes it's fascinating very beaut very american very the street finds its own uses except the street here is a giant acid metal bathtub and it's a giant acid metal bathtub next to an active open pit mine that's being currently worked on which is next to a let's make more than one of the most beautiful historic looking downtown cities that has been largely unchanged since the mining heyday yeah it's crazy sad that scarecrow job sounds pretty fun though well yeah you could sit in the middle in a house in the middle of an acid lake especially if you're a birds like person right if you're a bird person well speaking of birds jeff jarvis's pick of the week well i could do that one but instead i'm gonna just oh i gave you the perfect lead in all right go ahead i said i'm very proud of my pick today and it is cut your pepe puffs oh are they good good i don't know we're gonna find out right now we're going to have an unbegging so now if you have a memory you may remember many years ago jeff retreating from this show to go have his trader joe's cacio e pepe uh which is actually pasta with pepper and it's it's mac and cheese all right so this is um cheese let's see four and a half servings per container 150 calories totally you want to get your uh app and and check this before we he eats it yeah no do you want to uh take a big no you gotta crunch into it right in the microphone jeff you gotta give the people what they came here for

3:01:39there's nothing you know what noise cancellation is too damn good we don't hear anything is it tasty does it taste like it's a little weird there's there's all the ingredients i don't know if you see anything i mean i don't know if you uh these cultures yeah If you ever saw what they make, these things. Now, mind you, I could eat the whole damn bag, but no, I don't really saw.

3:02:09I saw a video of Cheetos composition somewhere. Let me see if I can find it. Inside the Cheetos factory. okay it's in 4k how do they make them puffy this looks like ai this does look like ai doesn't it yeah well anyway we start with prime kernels of corn this is ai you think this whole thing is ai yeah like they're harvesting cheetos oh yeah waste manage it up and even the text is wrong yeah uh well is there anything it can't do there's a happy guy buying cheetos and so this daughter can run home and eat that and hostess donuts and uh completely incorrect bag all the text is misspelled this is uh you know what we're gonna laugh someday at the early era ai videos i wonder where that went what the year of this was hey if you want to watch an hour and a half version of this millie check out works in the cheetos factory uh this is from a uh a youtube channel has four seven subscribers and it was posted seven months ago right now more people just watch that video right now on the screen than ever before so don't take us down okay man we're helping you out ladies and gentlemen that concludes this gripping edition of intelligent machines mr cacio epepe there is jeff jarvis and he's the author of hot type he's still eating them he hasn't keeled over yet he seems happy

3:03:57consumer reports her next story will be why you don't want to eat catch your pepe i'd never come for catch away pepe cheese and cheese is always good it's true uh we do the show every wednesday right after uh windows weekly that's 2 p.m pacific 5 p.m eastern 2100 utc you can watch us do it live in the club to discord or or also youtube We join the seven other people watching. No, that's more than that. YouTube, Twitch, Facebook. There are hundreds of people watching on YouTube. Hundreds. LinkedIn, Facebook, and Kik. Otherwise, you can download a copy of the show, audio, and video from the website, twit.tv.

3:04:39You could also find it on YouTube. There is a YouTube channel dedicated to intelligent machines. And then the best way, easiest way, subscribe on your favorite podcast client. you'll get it automatically on a Wednesday afternoon or evening after we're done with the show. Thank you, Paris. Thank you, Jeff. Thank you, everybody, for joining us. We'll see you next time on Intelligent Machines. Bye-bye. Hi there. Leo Laporte here. I just wanted to let you know about some of the other shows we do on this network. You probably already know about This Week in Tech. Every Sunday, I bring together some of the top journalists in the tech field to talk about the tech stories It's a wonderful chance for you to keep up on what's going on with tech, plus be entertained by some very bright and fun minds.

3:05:24I hope you'll tune in every Sunday for This Week in Tech. Just go to your favorite podcast client and subscribe. This Week in Tech from the Twit Network. Thank you.