In short
Big Technology Podcast - Episode Summary
Episode Title
Sam Altman’s Reflections, NVIDIA’s Robotics Play, Zuckerberg’s Moderation
Episode Overview In this episode, host Alex Kantrowitz and guest Ranjan Roy from Margins discuss the latest developments in the tech world, focusing on key statements from Sam Altman regarding AGI (Artificial General Intelligence), NVIDIA's ambitions in robotics, and Mark Zuckerberg's new content moderation policies at Meta.
---
Key Discussions
- Sam Altman on AGI
- Altman's Declaration: Sam Altman claims that OpenAI is now confident it knows how to build AGI as traditionally defined, with expectations of AI agents joining the workforce by 2025.
- AGI vs. GPT-5: Discussion on whether AGI could be declared before GPT-5 is released, with skepticism about the definition of AGI being lowered for fundraising purposes.
- Critique of Current Definitions: Roy and Kantrowitz express concerns that Altman's definition of AGI is vague and may simply be an expansion of what AI agents can do within existing workflows.
- Superintelligence Horizon: Altman hints at the future focus on superintelligence, which could propel significant advancements beyond current capabilities.
- Anthropic's Fundraising
- New Capital Raise: Anthropic is in talks to raise $2 billion, which would triple its valuation in just a year.
- Investor Sentiment: Despite many AI startups losing money, there’s significant investor interest in generative AI, but concerns about the long-term profitability remain.
- Strategic Positioning: Discussion on Anthropic's quieter approach compared to OpenAI, focusing more on building than on market noise.
- NVIDIA's Robotics Ambitions
- Multi-Trillion Dollar Opportunity: NVIDIA’s Jensen Huang declares robotics as a significant growth area, presenting AI models for humanoid robots and partnerships with companies like Toyota.
- Personal AI Supercomputer: NVIDIA announces a personal AI supercomputer priced at $3,000, enabling users to run complex AI models locally.
- Smarter NPCs in Gaming: NVIDIA is developing AI-powered non-player characters (NPCs) that can respond and interact dynamically in video game environments, potentially enhancing user experience significantly.
- Mark Zuckerberg's Moderation Policies
- Shift to Community Notes: Zuckerberg announces a move away from fact-checking to community-driven content moderation, similar to Twitter's approach.
- Policy Simplification: Meta is reducing restrictions on sensitive topics, promoting a more open discourse but raising concerns about the potential for hate speech and misinformation.
- Political Engagement: Meta is reintroducing civic content to increase engagement, which may open the door for politically charged discussions and potentially volatile interactions on the platform.
- TikTok Ban and U.S.-China Relations
- Zuckerberg's Timing: The announcement of Meta's new policies coincides with ongoing discussions about a potential TikTok ban in the U.S., with Zuckerberg potentially positioning Meta favorably amid political shifts.
- Supreme Court Hearings: Discussion on the Supreme Court's stance on TikTok and ByteDance, with justices questioning the justification of TikTok's claims against divestiture.
- Personal Reflections on China
- Alex's Experience in China: Kantrowitz shares insights from a brief visit to Beijing, highlighting the pervasive surveillance and cultural contrasts.
- Cultural Appreciation: He reflects on the importance of travel for understanding different societies, emphasizing the need for context beyond what media typically present.
---
Key Takeaways
- AGI Definition Dilemma: The tech industry is grappling with the definition and implications of AGI, raising ethical and practical concerns.
- Investor Optimism vs. Reality: There is a dichotomy between rapid investment in AI and the actual profitability and sustainability of many startups.
- Shifts in Content Moderation: Zuckerberg's approach signals a significant change in how platforms may handle misinformation and hate speech, which could reshape user experiences.
- Technological Innovations: NVIDIA's initiatives reflect a broader trend towards integrating AI into everyday applications, from robotics to personal computing.
---
Conclusion This episode provides a comprehensive overview of significant trends and developments in the tech industry, focusing on AI's future, content moderation strategies, and geopolitical implications in technology. The discussions prompt deeper reflections on the ethical and societal impacts of these advancements.
Written by AI. May contain mistakes. Listen to the episode to check what was said.
Transcript
Automatic transcript. May contain errors.0:00Sam Altman says OpenAI now knows how to build AGI. Why? Anthropic is raising another$2 billion. NVIDIA looks to a robot future. And Mark Zuckerberg says to hell with fact checkers. That's coming up on a Big Technology Podcast Friday edition right after this.
0:19You're used to hearing my voice on the world bringing you interviews from around the globe. And you hear me reporting environment and climate news. I'm Carolyn Buehler. And I'm Marco Werman. We're now with you hosting The World Together. More global journalism with a fresh new sound. Listen to the world on your local public radio station and wherever you find your podcasts.
0:46Welcome to Big Technology Podcast Friday edition, where we break down the news in our traditional cool-headed and nuanced format. We have a great show for you today. We're talking about Sam Altman's new bold statement about AGI. We're also going to cover the latest NVIDIA news, the latest Anthropic fundraising, Mark Zuckerberg talking about how content moderation is no longer as important to Facebook as it has been in the past and what that all means, whether it's just a play to get in the good graces of the Trump administration. And, of course, a little discussion of my visit to China and where the country stands today and where it's heading.
1:23Joining us as always on Friday is Ranjan Roy of Margins. Ranjan, great to see you. What better way to continue our 2020 Thrive than having Sam Altman tell us AGI is here, at least AGI-ish. I don't know about you, but I've gotten a lot of 2020 Thrive texts over the past week. To stop it. So I want to thank you for introducing that to big technology and my life. Thank you, Ranjan. I don't think my wife and family, I think, are definitely over me saying 2020 Thrive, but I'm sticking with it. I will admit I've been saying it as well. when someone talks about how this year is going off to a bad start.
1:59I'm like, just 2020 Thrive. Don't worry about it. 2020 Thrive. It's so easy. Manifest. All right. So someone who's manifesting is Sam Altman. He's talking about AGI. He wrote a very interesting post called Reflections, just looking back at the last two years since ChatGPT has released. So basically saying, all right, it's new to the new year, and I got to look back. We're going to talk about a few things that he's written, but to me, the most interesting statement, one that got the most attention this past week is this. We are now confident we know how to build AGI as we have traditionally understood that.
2:32We believe that in 2025, we may see the first AI agents join the workforce and materially change the output of companies. We continue to believe that iteratively putting great tools in the hands of people leads to great broadly distributed outcomes. I think that's a way of him saying that they're going to declare AGI this year. It got a lot of pushback in the popular press. I'm curious what you thought seeing that from Sam. I mean, I got a lot of pushback from me as well in my mind because it was just ridiculous. It's again, at least I think we were correct that at a certain point, he's just going to say AGI, he's going to move the goalposts to define it however he wants.
3:16To say that AGI is the first AI agents joining the workforce, in quotes, to me, makes absolutely no sense. Because agentic AI, and we have discussed this, we can definitely dig more into it, but it means something completely different, at least in my mind, and at least in the way most of the industry defines it, than how we've thought about artificial general intelligence for years. Because, again, AI agents are simply using large language models to take some autonomous decisions in some kind of existing workflow or business process. It's not that revolutionary or complicated. It's just some kind of like taking something that used to be rules based and letting an LLM try to apply a bit of logic or reason to it.
4:05That is not, at least in my mind, the robots are taking over or we have developed some kind of super intelligence. If he starts saying super intelligence is now different than AGI, maybe I'm okay with it. But overall, this read like, we need to raise more funds to me. Well, okay. So I think there's something that I agree with in what you said and something I disagree with in what you said. I agree that he's moving the goalposts. I don't necessarily think that he's saying that agents are going to be AGI. It might have just been that he's like stringing these statements together. But he did talk to Bloomberg about what he thinks AGI is.
4:46And this is the way that he defined it. And to me, it seemed like kind of a lower bar. He said, the rough way I try to think about it is when an AI system can do what very skilled humans in important jobs can do. I'd call that AGI. He says, then there's a bunch of follow-on questions like, well, is it a full part of a job or only part of it? Sorry, is it the full job or only part of it? Can it start as a computer program and decide it wants to become a doctor? Can it do what the best people in the field do or the 98th percentile? How autonomous is it? He says, I don't have deep, precise answers there yet.
5:18But if you could hire an AI as a remote employee to be a great software engineer, a lot of people would say, okay, that's AGI-ish. So to me, it seems like he's saying, okay, basically, if we're not quite there yet, we're almost there. AGI-ish. That is the term of the week of the month of the year, I think. To be able to say that it's kind of there, it's kind of what we've always promised, but we're just going to check off that box, I think kind of captures this perfectly. Again, like replacing specific functions in the workplace, we're already there. That's what I don't understand. Like GPT-40 is good.
6:04Gemini 2.0 Flash Experimental is good. Most of these other foundation models can do a lot of what people already do. So, you know, this idea, can it start as a computer program and decide it wants to become a doctor? it's really interesting how he sprinkles in these kind of fantastical statements within more just kind of monotonous like mundane things like yes it can do a human's job slightly better than them or a skilled human's job or it could suddenly decide it wants to become a doctor I think this is it's so tough to me that I don't know how how he's approaching this I like it was a nice post I like this idea of like taking some time to reflect.
6:51I think it was an honestly like a genuinely intentioned post that it's been two incredible, crazy years. And I want to look back on it. But still, just this whole, again, it's very Sam-ish in the way he's approaching this. This is from AI skeptic Gary Marcus. At a conference yesterday, someone with very good knowledge of open AI said something fascinating. More for what was not said than what was said. What was said is that we should expect to see GPT 4.5 soon. What was not said by the well-informed source was that we should expect to see GPT 5 anytime soon. Is it possible that OpenAI says we've reached AGI before GPT 5 comes out?
7:31Yes, 100%, 1 ,000%. He's setting it up for that. I think, again, we had somewhat said it in jest about him just kind of saying, OK, AGI is here, Microsoft contracts null and void. I think they're completely setting it up for that. I think they recognize that to release GPT-5, if it's not some massive step change in terms of ability, would be a huge problem. So I completely believe they're going to say AGI. I'm going to say by the springtime, by spring, maybe April, May, as we leave the cold months, we will have AGI. That sounds right to me. Maybe sooner. And then I think the discussion is really going to turn towards superintelligence.
8:22So I think the past two years, people have been talking about AGI. I think that they're going to declare, OpenAI is going to declare we've reached it or they've reached it. and then super intelligence is going to become the new buzzword and altman also previews this he says we're beginning to turn our aim beyond agi to super intelligence in the true sense of the world of the word we love our current products but we are here for the glorious future with super intelligence we can do anything else super intelligent tools could massively accelerate scientific discovery and innovation well beyond what we are capable of doing on our own and in turn massively increase abundance and prosperity.
9:02This sounds like science fiction right now and somewhat crazy to even talk about it. That's all right. We've been there before and we are okay being there again. Abundance and prosperity. Our super intelligent future is just one of greatness for all. I think we're seeing it happening in real time right now. This distinction between AGI and superintelligence, which were conflated, and now lowering the bar of AGI. And then I think that fantastical futuristic vision will be labeled superintelligence going forward. And it's savvy. It's smart. It'll help on the fundraising side. It'll help keep just kind of the general overall narrative and dream around AI in the overall media.
9:52But I think we're very clearly seeing it happen right now. Ranjan's like, just give me an intelligent dashboard. And Sam's like, we're building super intelligence. Yeah. Well, no, I mean, okay. So I said it last week, I'm going to say it again. My theme for 2020 Thrive is there is such a disconnect between what the industry is saying and what actual customers want. Because the idea that any fortune 500 business that has infinite complexity and systems and not clean data and relatively unstructured data is going to allow autonomous processes into their actual business anytime soon and allow it to be completely autonomous is a pipe dream in my mind.
10:39I think there's a lot of companies and a lot of steps along the way that will be able to take existing LLM technology and start doing incredible things. But I think like there's still such a disconnect. Everyone I talk to, like normal, even business technology people are not asking for totally autonomous agentic AI. They're just asking for stuff that makes life a little bit easier. So I think like he's I did one thing in the reflections. I also liked he did mention that they were a research house who kind of like almost backed into a business, which is something we've talked about a lot. And it's even more clear to me that OpenAI is still a research house that has a business tacked onto it.
11:27And I don't think that's the company that's going to win. Interesting. Yeah, even in the Bloomberg conversation, he talks about how they've had to protect research from like the standard wants of Silicon Valley style company and they have them in a separate location. So I think that's true. And one more thing about this, before we move on, So I want to just go back to this statement. We are confident we know how to build AGI as we have traditionally understood that. What do you think that path is? Is it just to continue scaling this stuff? Or, I mean, it's kind of interesting that he winks at it, but he doesn't exactly say what to do.
12:03This is, again, where I think they've been incredibly savvy. But again, the ARC AGI test that supposedly gave an 87.5 % score to the latest 03 model that's still private but being tested, that gave them enough credibility in the conversation to start saying we are close to AGI. but it's one of those things that even within the industry, a lot of people say that this is not actually an accurate test for kind of the traditional understanding of what AGI would be. But they have been able to say, look at the leaps and bounds of development we've done over the last just two years based on this one test and this one benchmark.
12:49So we're there. We know how to do it. It's done. We're good. Check. I mean, that's amazing. We'll see if this continues. One of the greatest salesmen of all time. I'll give him that. I don't want to spend too much more time on his reflections, but I did think that, as you mentioned, it was like a pretty good post talking a little bit about what's happened and the messy parts of it. And he's talked about his firing in this post, which I don't think is really worth rehashing, but I think he really just sums it up at the beginning. Building a company at such high velocity with so little training is a messy process.
13:17It's often two steps forward, one step back, and sometimes one step forward and two steps back. Mistakes get corrected as you go along, but there aren't really any handbooks or guideposts when you're doing original work. Moving at speed in uncharted waters is an incredible experience, but it's also immensely stressful for all the players. Conflicts and misunderstandings abound. I mean, maybe this is also just clever marketing from Sam, but even if it is, I just don't think you see this from high profile CEOs enough, like being like truly honest about what it's like to build a company and not just being like, you know, full systems, go ahead.
13:52And we are like taking over the world. Yeah, which is why I liked the post. I thought it was genuinely, I thought it was reflective. So being called Reflections was accurate. I thought he looked at it and it is. This is why I don't want to take away the technological marvel that they have delivered and kind of unleashed into the world over the last two years. It is something incredible. So to stop and take a moment to recognize that, that's the part I like. I wish we could just take a little more time to reflect on that and use the technology we have currently rather than having to switch instantly into what's coming next.
14:33So speaking of which, Anthropic is back raising more funds. They just raised a bunch last year. I think there is$4 billion last year. And they're going to try to raise another$2 billion. This is from the Wall Street Journal. Anthropic is in advanced talks to raise$2 billion in a deal that would value it at$60 billion, more than triple its valuation from a year ago. It's being led by Lightspeed Venture Partners, and it would make Anthropic the fifth most valuable U.S. startup after SpaceX, OpenAI, Stripe, and Databricks. This is from the journal story. Investors are excited about the potential of generative AI to transform how people work and live and are largely unconcerned that most AI startups are losing money because of the high costs of the technology and intense competition.
15:16competition. I mean, sort of an interesting statement in there about like the concerns about being profitable from investors on this. I guess that's how it always works. But still, like, just to have this blanket statement that investors are unconcerned, seems wrong to me. That being said, Anthropic hasn't really made a lot of noise recently. I think it's just been quietly building and we don't see these posts from their CEO, Dario Amadei about, you know, reaching AGI the way that you see it from Sam Altman. He does work a little bit more quietly than the counterparts at OpenAI, but I'm curious what you think this means that Anthropic is going to raise another$2 billion, going to be valued so highly.
15:59It does seem like that amount today is something that an AI research house will blow through in about a quarter, but it is significant nonetheless. Well, the$60 billion is the part that's almost terrifying to me because how do you like work your way into that valuation? How fast do they have to grow their revenue when you're tripling your valuation from just a year ago? I think this is where I will say like what you said around how Anthropic is basically doing this quietly. I like, I respect, in a way it's almost nicely strategic that they're kind of riding on the coattails, let Sam make all the noise and just kind of be the hype man of the industry and then quietly build.
16:48And Claude is, other than its usage limits that you hit even as a pro subscriber, is an incredible product. I think it's probably my most used generative AI product. So they, I think the way they're approaching it is very smart. I think all of these companies, is because what do you do at$60 billion? Are you IPO-ing? I mean, when you're trying to put together an S1 in public-ready financials for a company that probably is not going to look too good, you're not IPO-ing anytime soon. I think, do you just raise more money at a higher valuation? Do you hope for an acquisition at that scale? It's getting more and more difficult.
17:28Remember, when Anthropic at$8 to$10 billion, plenty of conversation around a lot of buyers at$60 billion And that number starts to dwindle a little bit. Yeah. And I think that there's going to have to be a couple of really large flame outs, given the amount of money that's been spent and raised and the amount of money you need to make to justify that. And I'm starting to worry a little bit about Anthropic. And I wasn't worried last year, but this year I am. I mean, you think about OpenAI. They, as we've said, after looking at their financials, have gone. they are planning to make the most of their money from ChatGPT.
18:06ChatGPT, as Sam Altman noted in his post, hovered at 100 million users for a while and didn't seem like it was going to increase that much. And then all of a sudden, it triples. It goes from 100 million to 300 million users and has all this new interesting functionality, including the voice stuff. And I think that that's important. But Claude is not catching on with consumers in the same way, despite being what I would argue is a better product outside of the fact that it lacks the voice capabilities. And so where does it go from here? Is it able to build a$60 billion business on API alone despite the fact that like some of its best features are the fact that it can talk to you in a more human-like way and a warmer way in a way that people have built these relationships with Claude as I spoke about with Casey Newton a couple of weeks ago.
18:54So I do worry a little bit about Anthropic. Is that unfounded? While I said I worry about the valuation, I actually am not worried about that breakdown. Again, so revenue estimates, OpenAI is estimated around 27 % of its revenue is on the API slash enterprise side and 73 % is on the ChatGPT side, which would include ChatGPT Plus. Whereas Anthropic, 85 % of its revenue is on the API side and only 15 % is you and me paying for a Claude Pro. So I think they are taking the bet that people building on their foundation models is where they're going to make money. And I actually think that's a smarter bet because I do think at a certain point, the consumer side of this gets more and more commoditized.
19:49I think Gemini and Google and kind of entrenched Microsoft and Copilot on the what is my chatbot that will help me and be up on my screen throughout the day for the average consumer. That's going to get kind of commoditized away into existing products, into existing ecosystems like Google or Microsoft. So I think I think Anthropic actually is taking the smarter bet versus OpenAI in this case. I would not be shocked if Anthropic comes out with a thousand dollar a month unlimited use version of Claude. The same way that OpenAI came out with its$200 a month. Make it a hundred, I'm in. 150 maybe. Because I hit that Claude usage limit way too much and I get so mad.
20:36Right, but Altman said that they're losing money on the ChatGPT unlimited subscription even at 200 a month. That was such an interesting thing to say. Again, ChatGPT, what's it, Pro or Plus Plus or whatever, the$200, whatever the branding they put on it. When you're saying on a consumer-facing product, and again, if you're paying$200, obviously it's still going to be more business usage, but still, on an individual per seat product, you're losing money at$200 a month. that. And to say that out loud to investors when you're going to be asking them for more money, it's baffling to me. But it almost, again, it still adds to this mystique that the amount of compute and power required for these services, there's something magical that none of you or us understand.
21:31That's the only reason I can see someone ever saying that when you're running a business that's trying to raise money. I think one of the things he might have been trying to touch on is the fact that this is so valuable to some people. And maybe that means there's room to raise the price even more. That's why I think Anthropic is going to quietly take notes and release the$1 ,000 a month version of Claude. But we'll see. What would you call it? We got Claude Pro. Claude 1000, man. Claude 1000. And then Claude 2000 coming soon. Claude 2000 for super special usage. Yep. Okay. So speaking of major opportunities, we had NVIDIA at CES this week talking about robotics and calling it a multi-trillion dollar opportunity.
22:12And Jensen Wang was out at CES, which is like the Super Bowl for NVIDIA. And he's talking about how the robotics is going to be the next stage for the company's growth. He said he announced a new range, this is according to the Financial Times, he announced a new range of products and partnerships in the physical AI space, including AI models for humanoid robots, and a major deal with Toyota to use NVIDIA's self-driving car technology during a keynote speech in Vegas. So to me, I think this is important because we've been talking about how AI can hit a wall with LLMs and what comes next. And one of the potential things that comes next is real world understanding.
22:52So instead of just trying to understand the world through text and maybe video, it's AI models getting out there in the world and understanding how they interact with the world that we live in. And that can be done through humanoid robotics, which, you know, can understand and perceive and plan, you know, effectively a way through the world in a way that you just simply cannot do with a text model. And I found it very interesting in this moment where we're talking about finally hitting AI, maybe AGI, finally hitting a wall with scaling outside of that, that NVIDIA potentially has this new way forward in robotics.
23:29What did you think about the new Tron John? Man, can we appreciate what we have today for a moment? Again, now, even Jensen going and having to go to the next big thing, and I get you got to go to the next big thing. But to me, I do think if someone's going to crack it, I think NVIDIA would be potentially the one. I also have to say, I love NVIDIA's branding on the way they, we just came up with Cloud 1000. Maybe it's good. Maybe it's not. But so they have, they're going to release a new computer called Jetson Thor. Jetson Thor. I mean, just like a mishmash of futuristic things. And then the robots will be powered by Groot, which stands for Generalist Robot 00 Technology.
24:21I mean, basically Jetson's Groot just pulling in all these kind of like sci-fi references and Thor as well. And they're paired. So I got to give it to NVIDIA for just making it a little fun, making it exciting, making it memorable. um i do think yes in the medium term this kind of idea of like robot robotic progress real world ai physical ai how you get things to interact in the physical world it's gonna be huge self-driving is kind of the first real manifestation of that and that's real we've both ridden in a waymo It's very real. What are the other applications of that? Again, in warehouses, Amazon for years has been able to use robots.
25:12There's lots of examples of really special purpose robots already doing things in lots of situations. So I think it's a massive opportunity. I just wish it was not what we were talking about right now. We could all just try to build our first agent or two. Right. But NVIDIA, the whole reason why NVIDIA has gotten to the place it is is because it has been thinking two steps forward and everybody else has been stuck in the present. Yeah, but they never said it before. Their genius was, we have a good video game graphics card. oh, by the way, we're not going to tell you. And suddenly you're going to all realize that the entire AI revolution is going to be powered by something that only gamers would get excited about before.
26:02Like they didn't go out and just say it, they just did it. So I think that that's the part to me that feels different now than before. You know, what's interesting for me was to hear Jensen talk about it as a multi-trillion dollar opportunity as opposed to something scientific, right? To me, is that like, is this just a play to the stock market? Like, do you think that's what's happening? I mean, reading about it in the Financial Times, pretty interesting. Yes, 100%. But also, I'll give, like, we're talking about OpenAI is a research house with a business tacked on. Jensen Huang is a business person.
26:38Like, I mean, he, at his core. So he's a, he knows what he's doing. So I think on that side, and I also, I don't see him as someone who, I think he, He's very savvy about signaling on this kind of thing, introducing this into the conversation. But I do have the confidence that there is a lot of real stuff happening if he is saying it that really could make this a reality. Yeah, I'm pumped about it. I mean, we're going to start to see so much more advances here, especially because with robotics, you can now just like run simulations with robots. and then do that in like the virtual world and then bring that to the physical world.
27:20And we're starting to see some pretty cool stuff with robotics. You know, speaking of the present, there was also some interesting news that NVIDIA is also releasing a personal AI supercomputer with its latest and most powerful AI chip, Blackwell, which is going to allow researchers and students to run multi-billion parameter AI models. This is from the Financial Times locally rather than through the cloud. And it's going to be available at the initial price tag of three grand, which I think to me, I mean, the Blackwell chips, what they usually go for 40 ,000. So the fact that you can get your own personal supercomputer for 3 ,000 and run your own experiments on it, I think is pretty cool.
Read the full transcript
27:57Yeah, I actually, I thought this was more interesting than the humanoid robot stuff because one, I saw it and I kind of want it and I don't even know why I want it. So it looks kind of like a souped up Mac mini and just somehow it'll just empower you to do amazing things. But also, NVIDIA, to me, 3000 is still a consumer-ish product. They say it's for students and researchers, but, like, I mean, the Vision Pro was 3000, and that's a pure consumer product. So to me, this could be quietly, I mean, obviously graphics cards were very consumer-focused, but they might be entering the consumer market it again.
28:43Yeah, I did a story recently about how universities are like all out of chips and basically cannot compete at all. And to me, this seems like a bit of an antidote to that. I mean, I think that NVIDIA really saw the need there and decided to go for it. And I think this is really positive news. Yeah, I think anything NVIDIA does, they still have, you know, kind of the magic touch that until they really flop on something, I think you have to just buy into whatever they're selling right now. Last bit of Nvidia news that I'm actually personally pretty excited about. Basically Nvidia is building these souped up NPCs or non-player characters, non-playable characters in video games.
29:25And to me, this has always been like one of the dreams and I really regret not writing about it earlier because it's something that's been so obvious and in development. And that is that when you're playing video games, There are these non-playable characters that basically just kind of stand there and they're just like clearly like dumb video game stand-ins. And the promise with AI is they can actually become much more human-like and that there's no more such thing. There's no such thing as an NPC anymore because every, you know, bot within a video game can be intelligent. And it looks like NVIDIA is also working to power some of these new characters.
30:01This is according to The Verge. These characters that can use AI to perceive, plan, and act like human players. This is according to an NVIDIA blog post. Powered by a gendered AI, this new technology will enable living dynamic game worlds with companions that comprehend and support player goals. And enemies that adapt dynamically to player tactics. The characters are powered by small language models that are capable of planning at human-like frequencies required for realistic decision-making. as well as multimodal small language models for vision and audio that allow AI characters to hear audio cues and perceive their environments.
30:37I mean, I'm just thinking about running around in a video game and realizing that every character there is like, you know, quote unquote intelligent and has its own personality and feels a lot more like the real world. I think this could be really big for video games. I think this could be big for virtual reality. Heck, maybe it even brings back the metaverse. Metaverse is back. Metaverse is back. We populate it with AI, human-like AIs. That sit side by side with actual human players. I think this was pretty cool and it was very interesting. Yeah, I like, see, this is the kind of stuff versus the medium-term futuristic humanoid robots.
31:14This is the kind of stuff I like to see. This is, because this really, I have no doubt is happening and will actually just like unleash really cool things and experiences just in the next like year or two. I also noted my other, I think 2025 prediction. So they use small language model SLM. I've been seeing a lot of people in the agentic space using LAM's large action models. I think everyone's gonna start rebranding LLM into other kinds of terminology. technology realizing llm branding is kind of a little played out a little uh little restricted to wait when's gpt5 coming so i i think i think you're going to see every one of these companies start to come up with these alternative terms that are a little more focused slm lam things like that yeah the branding and ai it's just uh it's all branding it's all branding okay so one last bit of AI news here and sort of the dark side of AI, which again, we're constantly reminded that so much of this AI revolution is built on copyrighted materials.
32:25There's a TechCrunch story that says Mark Zuckerberg gave Meta's Lama team the okay to train on copyrighted works. This is coming out in court filings that Mark Zuckerberg cleared his team to train on LibGen, which is a data set. This is according to internal messages, a data set we know to be pirated that may undermine Meta's negotiating position with regulators. And when that concern was brought up, this is according to the filing, the decision makers within Meta said that after escalation to MZ, Mark Zuckerberg, the AI team was approved to use LibGen. I get the argument that this is transformation of people's work.
33:10I do not fully agree with this idea that therefore companies can go ahead and ingest copyrighted works and use them freely for their purposes without a license. I don't know. It just seems wrong to me. What do you think, Rajan? Well, we're going to be talking about Mark Zuckerberg's announcements regarding the meta platforms in just a moment. And I think there is going to be some significant copyright lawsuit resolution of some sort in 2025, like, you know, the New York Times suing open AI already. When stuff like this comes out, it's one of those that I really think from like, I like that this whole episode is basically AI is all branding and PR.
33:56But I genuinely believe that like you need to have the public on your side when this stuff comes out, because on one, it's so crystal clear that obviously there's copyright violations in all of this. I think everyone, it's hard to pretend that that's not the case. However, these products are so valuable to everyday users. It's like that when everyone supports you and is on your side, I think it's just okay enough that these things will resolve themselves. But I think like if the general public is turning against you, I think this is where there can be some real issues going into the next year.
34:39Because I think we're going to see more and more of these lawsuits. And at some point, there has to be precedent set, because this is completely uncharted territory. And we are going to, there's going to be some kind of court resolution that establishes some kind of precedent. Yeah, no, I'm with you 100%. I mean, ultimately, I get the argument that the AI companies are making and why they'd say it's fine to train on copyrighted material. But there's just something not right about it. And so far, there's been, you know, a small public backlash, not really much of any. and I just think that that's going to come to a head eventually.
35:14Alright, so speaking of meta, there's some new moderation policies that when we speak about them we're sure to make at least half of our listeners mad, maybe all of them mad, and that we will not be daunted. We will discuss them right after this. Did you know your credit card points and miles can lose value to inflation? Credit card companies often reduce the redemption value of your points and miles. Now imagine a credit card with rewards that can grow in value. With the Gemini credit card, you can earn Bitcoin or one of over 50 other cryptos instantly with no annual fee. Every swipe at the store or gas pump earns you instant rewards deposited straight to your account.
35:54Plus, sign up now for a$200 Bitcoin bonus to kickstart your rewards. Visit Gemini.com slash card today. Check out the link in the description for more information on rates. Again, if you're looking to invest in Bitcoin but don't know where to start, the Gemini credit card makes it easy. The Gemini credit card is issued by WebBank. In order to qualify for the$200 crypto intro bonus, you must spend$3 ,000 in your first 90 days. Some exclusions apply to instant rewards in which rewards are deposited when the transaction posts. This content is not investment advice and trading crypto involves risk.
36:30The Gemini credit card may not be used to make gambling-related purchases.
36:37You're used to hearing my voice on the world bringing you interviews from around the globe. And you hear me reporting environment and climate news. I'm Carolyn Buehler. And I'm Marco Werman. We're now with you hosting The World Together. More global journalism with a fresh new sound. Listen to The World on your local public radio station and wherever you find your podcasts.
37:04And we're back here on Big Technology Podcast. First half, we talked all about the latest in AI. Now we're going to talk about the second biggest story, I think, in tech this week, which is a meta full about face on its fact checking and content moderation policies. So basically, Mark Zuckerberg says they're going to take a whole new approach to speech on Facebook. he basically says there are five changes they're going to make. One, they're going to get rid of fact checkers and replace them with community notes like X. Second, they're going to simplify content policies and get rid of a bunch of restrictions on topic like immigration and gender, which Zuckerberg says are just out of touch with mainstream discourse.
37:47Third, they've had filters scanning for any content policy violation. They're going to tune the filters to focus on tackling illegal and high severity violations, leaving the rest to user reports. Fourth, they're bringing back civic content which is a way to say that there's going to be a lot more news and politics on Facebook. And fifth, they're moving their trust and safety content moderation teams out of California and to Texas. And they're also, this is like really a sixth thing, they're going to work with President Trump to push back governments around the world that are going after American companies and pushing them to censor more.
38:26Basically, they've had other countries lean on them to censor, and they're going to try to work with the Trump administration to push back on that. So those are the big changes coming out of Facebook. There's more detailed stuff that we're going to get into, but just on its face, what do you think about these changes, Ronjan? I think Mark Zuckerberg has had an incredible PR turnaround over the last few years from back in the 2017-18 Cambridge Analytica days. I think he's definitely risking all of that work to make this public statement because there didn't need to be him saying this. That was interesting to me.
39:12like from a policy perspective, from an actual product perspective, they could have just done this to obviously it's so saying it publicly is kind of, you know, presenting this to Trump very clearly, kind of like presenting this to the world. But I think and I'm curious, your take my possibly countering to of opinion is, I think it's good in the sense that I have long been wary of misinformation on meta platforms. I think fact checking the adjust at that scale, or the any of these efforts that have launched over the last six to eight years, have been kind of terrible and a fool's errand anyways.
39:56So I kind of like just let it rip as awful as that sounds like it just this is what this platform is when you have an algorithm based content platform like this. This is the way it's always going to go. So to try to hide it and make editorial decisions or value-based decisions was always going to be a problem and now they're just saying sorry we're not it's not our not our problem yeah no i think going from fact checking to community notes is a good move uh i think community notes really works very well on twitter in fact i'm impressed with how well it works and go ahead i i okay sorry i i disagree on that.
40:37I don't think from an actual like efficacy standpoint that any of this is going to work well in terms of like creating a better town square conversation. I just think it's the the only real decision and it's actually the most honest one. All right but I'm yes and on this. I just think community notes is good. I mean it's amazing how many viral tweets that you used to see that would just go viral without any context now go viral with fact checks, even ads on Twitter, when the advertiser is lying about their product, they get community noted. I mean, and it's not like, oh, the community notes generally seem like pretty level headed.
41:16And oftentimes we'll have links to like, share further context. So why is this not a great solution here? Well, because, okay, so we have to separate, there's the kind of, there's the fact checking side, which to me was always kind of ridiculous anyways. Then there's the actual kind of like abuse side, which also they've like severely relaxed the overall restrictions of it, which again, if you say something awful, that's not wrong or right factually, that is just something that is very like, and we'll get into things like immigrants are grubby, filthy pieces of shit, which was one of the examples of now, now newly permissible speech.
41:58I mean, do we need a fact check? Yes or no? Like, so to me, the community note side, it does answer the misinformation slash fact checking side. I do agree. It's kind of, I love seeing an advertiser get called out with community notes. But to me, that's almost a small part of this. Well, I think that's a big part of it because it's on the fact checking. But then I think that you're right in pointing out that the abuse stuff is a completely different side of it. And that's again, going towards where Zuckerberg talks about how, like, we're going to simplify our content policies and get rid of restrictions on topics like immigration agenda that are just out of touch with mainstream discourse.
42:37He says what started as a movement to be more inclusive has increasingly been used to shut down opinions and shut out people with different ideas. And it's gone too far. And like, while I think that's possible, it's kind of interesting to see what's now permissible on Facebook. I mean, you read like a few things and there are some other things about trans people. Trans people aren't real. They're mentally ill. Trans people are freaks. It's very interesting to me how like where Zuckerberg says broaden out the discussion really is just like, let's get into like the most vile parts of our society and let them say the things that they want to say, which I guess like, yeah, you don't want to limit everybody's personal expression.
43:14Of course, I don't think any political opinion should be restricted. And I'm definitely not on like the pro heavy content moderation side on social networks. On the other hand, like you are building a platform that does amplify stuff algorithmically. And like you get to kind of set the parameters of like the type of world you want to build inside. And it was just kind of interesting from like, you know, it's like Zuckerberg says, let's broaden out the political debate. And this is from platformer. This is from a Meta's chief marketing executive. I feel that the actual shock of friends and family seeing me receive abuse as a gay man helped them understand that hatred exists and hardened support.
43:56Most of our progress on rights happened during these periods without mass censorship and pushing it underground has coincided with reversals. Like that is an interesting idea. But the question is like, how much of like the negative stuff do you want to amplify, right? And I think Casey Newton, who wrote that, wrote the story with this quote, pointed out pretty well, which is basically just like, Meta's moving to like a more of a for you feed, less of us, you know, people you follow feed. So the real question isn't exactly like, how much of your of this do you want to allow? It's how much of it do you want to amplify?
44:30Well, I but I also think that, in the end, the up until now, with Facebook and Instagram having relative kind of like social network monopoly positioning. This is the only thing that the average person like you and I are very online. I've dealt with plenty of abuse, especially on Twitter over the years. And it's fine. I, you know, like I am online, I deal with it, I know how to deal with it. Like, for just the average Facebook user, which I'm guessing Facebook blue has gone more normie than ever. How are these people going to respond when they start getting flamed with just vile remarks and just feeling horrible.
45:13Do they stop using the platform? Do they use it less? So I actually think, to me, and I kind of hope this resolves itself from an actual kind of product and business perspective. Maybe that's like overly capitalistic of me. But I do think that these platforms will become just miserable and unusable the more, again, for me, I know you're still anti-Blue Sky and still strong on X, but like, I mean. I wouldn't call myself anti-Blue Sky. I just don't think there's a big, you know, bright future there. Well, no, but like when I go on X now, it's wild. Like it's crazy. I mean, like the UK grooming stuff and just like, and I've been going on it less and less and less.
45:58And then when I go on, just the outright like insane racism, not even like kind of like subtle or even funny or whatever. I mean, just like that as well, just. And so, but the more that happens, the less I inclined to use it. And I mean, that's why I'm even think blue sky is a chance that I, for my own behavior, I use it more and more and more now because as the other alternative becomes less and less usable. So, so I do think it's going to be interesting to see with, to see where this goes from an actual business and product perspective, because it all sounds good right now, but is this going to make these things unusable for just a normal person who's just like posting about their kids or whatever else on Facebook and then just starts dealing with just insane, insane comments?
46:49Well, I did have an observation from the business side of this as well, because it's interesting to me that they're bringing back politics. I mean, the civic content thing, I think to me, is the most undercover part of this entire thing. And I think they know that news and politics give gives their products engagement they just can't find elsewhere. And so I'll sort of question whether like people flaming each other and blowing each other up in the comments is bad for Facebook's business. In fact, it seems like Facebook blue was at its height when the flame wars existed and people would go after each other and they'd be talking about news and politics and And whatever it is today, it just doesn't have the same urgency as it did before.
47:28So I think that like on its face, this is absolutely, and I think in truth, this is absolutely a move of Zuckerberg to align Facebook with the current political administration as it has always done. It always aligns itself with the political conventional wisdom in the country and I think across the globe. And Zuckerberg has said, point blank, our job here is to just reflect what people want. Like we don't have anything higher purpose than that. What people want, we give them. People said something in the election, Trump won, and therefore he's aligning himself with Trump. However, second order right now, I think is largely, it's a business thing as well.
48:07I think he realizes that to give threads a chance against Twitter and to restore urgency to Facebook, you need news. News equals engagement. And I've said it a thousand times on the show. When you have news, when you have politics, you have engagement, you have urgency. And I think what Zuckerberg is doing is bringing that back into the platform and, you know, and all the horrible things that come with it in order to revive what is, you know, a platform that seems to be on its way to irrelevance in Facebook and certainly there in threats. Well, I also think what he's doing, which is cynically savvy, is announcing this just a few days before.
48:48And today, I was January 19th, the deadline for TikTok needing to be divested from ByteDance. and listening today to the Supreme Court hearings about the TikTok ban, suddenly it went from Trump supporting TikTok again, being against a ban, to now Mark Zuckerberg is fully aligned with me. He sees it. He knows it's happening. And I mean, you can start to see suddenly, and I was listening to the hearings, like even the conservative justices seem to be edging towards a ban. So if Trump does not step in on this, I think this could actually be the thing that could push over the ban to a reality. Do you still see that?
49:31Do you see the ban happening? Oh, yeah. OK. So today, listening to the Supreme Court hearing, the most fascinating thing, I mean, in this day and age, the nine justices on the Supreme Court seemingly being aligned on anything, I think, is almost non-existent. they all seem to be leaning towards and again it's not a ban and actually the funny Amy Coney Barrett kept saying it but then also Justice Jackson on the liberal side was saying it they're all like this is not a ban it's a divestiture like all you have to do is divest from ByteDance why can you not from a you know a foreign adversary like a company why can you not divest And then the solicitor on the ByteDance slash TikTok side keeps saying this is a matter of free speech.
50:24And they're like, no, it's not speech, just divest. And it's the fact that everyone from both sides, even Clarence Thomas is like, what is exactly, what is TikTok speech here? I don't understand why a restriction on ByteDance, a Chinese company, represents a limit on TikTok. So I think Zuckerberg's move might actually be the defining push where Trump suddenly is like, OK, I don't have to worry about meta on Instagram right now. You know what? I can now go after China. Let it rip. Let it rip. Yeah, that's really interesting. And I think, again, like we've talked about this on the show, but if TikTok ends up just shutting down and not divesting, then that's kind of the game right there.
51:10In fact, this is probably a good test to be like, you know, are you going to try to divest or are you going to shut down? I still, heart of heart, don't believe that this thing is going to get shut down or sold or, you know, maybe there'll be some declaration of divestment and that'll be that. However, if it goes through and TikTok says, all right, we're done, it's like, okay, you are a tool of the Chinese Communist Party. But they've already said that. All they had to do, like to say, we are not going to use the ByteDance algorithm and completely divest our data, our algorithm, the business from ByteDance.
51:42That's not a completely impossible, unreasonable thing. Like, that's the part to me that still is almost like confounding. Like, why can't they just do that? And that's what you heard it even in the justices when they were talking again across the aisle. Everyone's like, why can't you just do this? Like, you're saying it's free speech. you're doing all that like it's not unreasonable that the foreign adversary of like an opaque company that we don't know exactly the relationship with the chinese government you should not be controlled by them when you control the narrative of american media like and they there's no good answer to that question why can't they divest again we're only nine days away.
52:27It's going to be a fun, crazy nine days on this one. I think we're going to get... All it's going to take is one Trump truth social. What are they called again? Truth social. To say that... What's the actual post called if it's a tweet? I'm not that far down. I still call X posts tweet, even though they call them posts. All it's going to take is one... Maybe they're called truths? I don't know. Truths. truths. That sounds good. One truth saying, let TikTok, let it shut down, and then it's over. And I think he's going to be leaning on that way, could be leaning in that direction, given Zuckerberg's announcement, given just overall, he's going to need anything that makes him look weak on China is always going to be problematic.
53:20And so this was always going to be kind of a delicate situation anyways, versus cleanly saying, I don't want China involved in anything, shut it down. Yeah. So speaking of China, should we close on the fact that I spent 15 hours in Beijing on Tuesday on a layover back home and just found it very interesting to be in China? You know, you've been there as well. To me, the thing that really stood out was the fact that it really is a surveillance state. I mean, there are cameras everywhere. And we were driving myself, my wife and a guide, we were driving out of the airport on our way to the Great Wall to start the day pre-dawn.
54:02And it just felt like every few seconds there was another flash. And I was like, oh, like in Australia, there's lots of speed cameras. And of course, I knew about surveillance in China, but I was like, is that like a speed camera? It's like, nope, That is what they call eagle eye or sort of the nickname for it. That just tracks everybody's movement, does facial recognition to see who's in the car. And if you do things that are wrong, they will get you. And in some places like in Tiananmen Square, every light fixture seemed to have like a dozen cameras or more affixed to it. And it's just this incredible thing that I don't think you fully grasp until you see it.
54:43that like wherever you go, wherever you go, you're being watched. It was very interesting to see in person. Well, I, so I spent three months in 2009 in Beijing and I like lived there. I was taking a language class and that, that just reminded me the surveillance. So 2009, not quite digital, but it's such a different mindset in the sense of, so it was around the time of swine flu. and so like already there was like a lot of pandemic-y behaviors going on and i went out i went out one night and i was a bit hung over the next day and i didn't feel like going to my class so i emailed the teacher alan i'm feeling a little sick i don't think i can make it today two health officials showed up at my door and i was like yeah yeah and like literally and i did they did not speak any English.
55:36I, my Chinese still is not very good. And at the time was non-existent and literally start asking me and yelling me, yelling at me, asking me all these questions. And that, that, that entire mentality of surveillance and like kind of a collective watching even in the pre fully digital era was there. So I can't even fathom what it is like right now. Yeah, it's pretty incredible. Like I, at one point I was like, oh, where's my phone? Of course it was in my pocket. And the guy was like, don't worry if you left your phone like a few steps back, you know, it'll still be there because the cameras will make sure that nobody took it.
56:14And if somebody takes it, we'll be able to find out where they went with it. Thank you. Thank you. Oh my goodness. But, um, you know, I mean, outside of that, I just think that it was like a really great visit, honestly, to be able to see China in person. And it's just like, it was pretty cool to see, you know, just a culture that has been, you know, ongoing for so many years, just reflected in the modern day. As you like walk around Beijing, you could really see it with the architecture and sort of like the dress. It was just super cool. And obviously, like getting a chance to see the Great Wall, which is just like one of the wonders of the world was really special.
56:52So I got a 10-year visa to go. And so 15 hours of those 10 years are in the books, and I definitely have to go back. Did you download Douyin, the original TikTok based out of China? I should have. I tried to download Alipay, and I did get it to work, but I still had some problems with it. Like, you can literally not pay with... You need a local phone number, right? Probably. I think that's what it is. Yeah, in Taiwan, I have to deal with the same thing with, yeah. Yeah, so couldn't use, could not use credit card. Everything is Alipay. There are so many EVs all around. Like you see the EVs with the green license plates, and there were so many brands I'd never seen before that are just driving around.
57:38It was almost like sensory overload to like see all these things that we had like, you know, read about, that I had read about in the press, like the surveillance, the EVs, the mobile payments, you know, the Soviet influence even, like you'd go from the Forbidden City, which is all ancient Chinese architecture, into Tiananmen Square. And then you like look around and there's like the Great Hall of the People, which is the most important building, I would say, in the country where all the government meetings are held. And it's like, oh, that's just like straight up Soviet architecture. And so it's like, look at the flag.
58:12You're there. You're there. It's real. Yeah. You're real. the soviet themes in the flag um you know you go to the market there are like sculptures of uh of uh ayatollah khamenei you're like okay that bond is real you know it's just pretty just a very very interesting trip and i did i think like i wrote this about this in big technology this week but there have been some debates on like whether you know it's worthwhile to travel and i just find those debates like so absurd because um not only do you get a chance to meet people from different backgrounds when you go but you also are left with so many more questions and a little bit more context of all these things that you see.
58:46And so I'm strongly on the affirmative side of go out and see the things that you want to understand. I'm not going to claim to be a China expert after 15 hours in the country. No, you should. You should. I think it's time. But it certainly opened a lot of new questions for me. And it was really, really helpful, I think, to see this in action. Are you going back? Yes. You got the 10-year visa. so I will go back as long as things stay kind of calm between the US and China I hope they do but yeah we'll see I might have to big technology live in Beijing not sure how that would play maybe we're not coming back I don't think Xi is going to allow that but it's a good idea maybe no maybe he's a listener maybe he's a listener I do know yeah that's one of the things that Xi Jinping enjoys.
59:38If you're out there, Xi, come on the podcast. I would love to have you on. Can you imagine our first head of state is Xi Jinping? I don't know. I think we have people asking questions.
59:50All right, Ranjan, great speaking with you as always. We'll see you next week. See you next week. All right, everybody. Thank you for listening. We're going to be talking replica on Wednesday, so that'll be a really fascinating conversation. And then Ranjan and I will be back next Friday. Thanks for listening. We'll see you next time on Big Technology Podcast. By the way.
From the publisher
Ranjan Roy from Margins is back for our weekly discussion of the latest tech news. We cover 1) Sam Altman declares the path to AGI is clear 2) Could AGI come before GPT-5? 3) Up next: Superintelligence 4) Anthropic raising $2 billion 5) NVIDIA says robotics is a multi-trillion opportunity 6) NVIDIA has a personal 'supercomputer' 7) Smarter NPCs are here 8) Meta's AI training copyright issues 9) Zuckerberg's fact check reality check 10) Motives of Zuckerberg's moderation moves 11) TikTok ban might actually happen 12) Alex's visit to China
---
Enjoying Big Technology Podcast? Please rate us five stars ⭐⭐⭐⭐⭐ in your podcast app of choice.
For weekly updates on the show, sign up for the pod newsletter on LinkedIn: https://www.linkedin.com/newsletters/6901970121829801984/
Want a discount for Big Technology on Substack? Here’s 40% off for the first year: https://tinyurl.com/bigtechnology
Questions? Feedback? Write to: bigtechnologypodcast@gmail.com


