Warren Buffett Compares AI to the Invention of the Atom Bomb

0:00On today's AI Breakdown, we're discussing why Warren Buffett sees AI like a new atom bomb, and also what the current state of the geopolitics of AI really is. Before that, the AI Headline News Brief covers new open source tools, updates from Google and Amazon, and a new text 3D model from OpenAI. What's going on, guys? Back with another AI Breakdown Brief. All the headline AI news that you need in five minutes or less. We're starting today with actually something that's not news so much as just a resource. Hugging Face last week launched their daily papers feature. It's huggingface.co slash papers.

0:37And this is basically a place where you can find all of the latest research coming out of AI. So they've got here avatar fingerprinting for authorized use of synthetic talking head videos. You can see the talking head generator, really cool stuff. And it goes on and on and on. This is just a great place to see what the bleeding edge is. So it's something that I've bookmarked and I wanted to make sure you guys could too. But with that, we move to our next topic, which is what big companies are doing to keep themselves fresh and competitive in the era of AI. We've heard a lot about how Google might be changing their search experience, a search experience that has been very consistent for the past 20 years.

1:18Project Magi is the codename for their search initiative. And it sounds like later this week we may actually be getting more information about that as they have their big developer conference. It's their IO developer conference that's happening this week. So apparently what we're going to see is a shift away in how they present their search results. There's going to be more short video, more social media posts. They're calling it snackable, which is a very marketing type word. But I think the main thing here, holding aside any of the specifics, is that there is just a new experience that people expect and a new type of user interface that's wrought in part by AI.

1:56Amazon, likewise, is looking to change the way that Alexa works. Now, on their recent earnings call, we heard their CEO talk about how they needed to build or they were in the process of building an even more powerful LLM to underlie Alexa. But it sounds like it's not just the technology underneath. It's what Alexa actually does that they're reconsidering. One example that they give is Alexa generating a bedtime story after an eight-year-old asks it to tell a story about a cat and the moon. If they develop that sort of creative capacity, it also opens up all sorts of new partnership opportunities like Disney.

2:30Again, I think the thing that's important here is less the specifics of whatever Alexa might do. And in fact, I think we're kind of over-indexing for the importance of personal assistance in the era of AI. But what matters is that Amazon is fundamentally reimagining its strategy on the basis of this new technology. Next up, we move over to the world of open source. Of course, one of the big stories from last week was that leaked Google researcher note about how open source is destined to beat Google and open AI and all of the closed approaches to AI. And we're seeing a lot more progress. I mean, it is coming hard and fast.

3:04We talked a little bit recently about Red Pajama, which is an open source version of Facebook's Llama, which if you're a parent, you'll get the joke. and they've just updated releasing new 3 billion and 7 billion parameter insight family of models that include base instruction tuned and chat models. So more to work with for developers there. We also had Mosaic jump into the space with their MPT-7B, a new standard for open source. They're calling it commercially usable LLMs. They say that it was trained on the Mosaic ML platform in nine and a half days with zero human intervention at a cost of 200K.

3:37And with this tool, developers can now train, fine tune, and deploy their own private MPT models. Speaking of people deploying their own models, Brian Rommel writes, with WizardVacuna M1, we now have an almost a chat GPT-3.5 running on your local hard drive on a simple modern computer. No internet needed after download. Personal private AI you own. In the space of six weeks, we have moved to the Cambrian explosion. Using chat GPT-4 as a judge, WizardVacuna LM is almost equal. This is exactly what that anonymous Google author was talking about, just an incredible rate of innovation from independent developers.

4:13Finally, I wanted to call out a new product from OpenAI called Shape E. This is Alex Nickel, one of the lead researchers and developers on it. He writes, super excited to release Shape E, our latest work on 3D generative modeling. So effectively, this is text to 3D modeling. And you can see the samples they have here are a chair that looks like an avocado, an airplane that looks like a banana. a spaceship, a birthday cupcake, a chair that looks like a tree, etc. This obviously has a ton of applications from VFX to gaming to 3D printing. Rowan Chung here says text to 3D printers are about to be a thing very soon.

4:49So now we have text to image. Increasingly, we have text to video and now text to 3D printing. This is obviously very nascent, especially compared to those other two, but it still seems like the trajectory is clear and people are going to increasingly be able to input their ideas in text and have them come to life in the real world. That's it for today's AI Breakdown Brief. If you're enjoying this, if you're finding it useful, please give this video a like or subscribe to the channel, and I'll be back in a little bit for the main AI breakdown. Last week, Warren Buffett likened AI to the atom bomb.

5:22Today on the AI Breakdown, we're talking about what he meant and what it says about the state of the geopolitics of AI. If you've been on AI Twitter in the last few days, you've likely seen some version of this headline. This one is from Business Insider and says, Warren Buffett says AI is like the creation of the atom bomb. He often calls nuclear war his biggest fear. So what happened was that last week was Berkshire Hathaway's annual meeting, which is a big, big meeting for capitalists in which one of the main events is Warren Buffett and his longtime partner, Charlie Munger, sitting and answering a set of questions.

5:59When one person asked him about AI, he had a very interesting response. And rather than just trying to sum it up or giving you the pull quotes that are most interesting to me, let's just listen to a big chunk of it in total. It can do all kinds of things. And when something can do all kinds of things, I get a little bit worried. And because I know we won't be able to uninvent it. And, And, you know, we did invent, for very, very good reason, the atom bomb in World War II. And it was, you know, it was enormously important that we did so. But is it good for the next 200 years of the world that the ability to do so has been unleashed?

6:52We didn't have a choice.

6:56But when you start something, well, Einstein said after the atom bomb, he said, this has changed everything in the world except how men think. And I would say the same thing may, not the same thing, I don't mean that, but I mean with AI, it can change everything in the world except how men think and behave. And that's a big step to take. A good question, and it's the best answer we can give. So there are two really interesting parts of this to me. The first is this idea that we can't uninvent this. I think this is the sense that once unleashed, this is a genie that can't go back in its lamp. It is toothpaste that can't be put back in the tube.

7:51It takes on an inevitable force. And certainly we've already seen that in the industries where AI has started to take hold. If you've ever used AI tools for creating or chat GPT for research, it's not hard to understand this no turning back kind of idea. Now, there's also an additional dimension of this, which is the competitive dimension, which on a company to company level, we're seeing more and more as these big tech companies hurdle themselves forward, trying desperately to keep up with one another or stay ahead of one another, and especially to stay ahead of competitors. This, more than anything else, according to Jeffrey Hinton, who left Google last week and has been on an AI safety tour ever since, is the source of his concern around why this has gotten so different.

8:41Why AI is now proceeding at the speed with which it is. It is no longer just a small handful of companies proceeding cautiously. Instead, it's one of the most significant capitalistic arms races that we've ever seen. And in that type of setting, precautions and cautiousness in general go out the window in favor of other objectives. And that, of course, I think gets to Buffett's second point, the idea that this won't change the way that people behave and think. Now, of course, there's lots of ways to interpret what he's saying, but my guess is that what he means is that we should be proceeding with perhaps more caution than we are, just given our track record of how we can use new technology, not just for good, but also sometimes for ill.

9:23And what that brings up for me is the question of and conversation around the geopolitics of AI. AI comes into a world that is in the midst of a geopolitical realignment already. We're moving from a unipolar, US-led world order post-Cold War to something that looks more like a multipolar world with various sources of power around the world, certainly with China and the US in the lead, but other areas including Europe and emerging powers like India, Brazil, and other places starting to exert their own dominance relative to the United States and other global powers. In that multipolar realignment, AI could potentially play a significant role.

10:02And you see this referenced constantly. On screen is an article from Bloomberg that's a warning from venture capitalist Vinod Khosla, who says that slowing US AI research would be hugely to the advantage of China. Kostler says we shouldn't take for granted we will lead in AI technology 20 years from now. Now I think you hear about this China threat quite a bit when it comes to AI. You see it as a way to justify a light-handed US regulatory approach. You also see China as one of the arguments against any sort of pause policy like the one that Open Letter recommended just a few weeks ago. The rejoinder that you'll sometimes hear is, listen, if we pause but China doesn't, then the pause didn't actually accomplish anything at all except letting China get out ahead.

10:43Now, going beyond the rhetoric around China, is there anything we can learn about their actual strategy vis-a-vis AI? It is certainly the case that it is a central and important issue. It was mentioned in a Politburo meeting a couple weeks ago, and then just last Friday, Chinese Premier Xi Jinping again urged the country to seize the opportunities of AI. Interestingly, this Axios article argues that China is far ahead of the U.S. when it comes to AI regulation, effectively because they've been dealing with it for years. The piece writes, After lagging behind the West through consecutive industrial revolutions, leaders in Beijing are driven by a determination not to be humiliated again in the AI era.

11:21Chinese authorities now have six years of experience building up AI regulatory know-how since they launched a next-generation artificial intelligence development plan in 2017. They're using regulation as a form of industrial policy in addition to traditional subsidies. Going on, the article also quotes Matt Shaheen, a Chinese AI governance expert at the Carnegie Endowment, who told Axios that, quote, Beijing's first regulatory efforts were stoked by a fear of AI's downsides, and it has learned a lot over six years. Shaheen said, AI regulation is highly iterative. It's about building up the muscles.

11:55There's very little chance that you're going to get AI regulation completely right on the first go. Now, while the thrust of this Axios argument is that the U.S. needs to be doing more when it comes to regulation, it's certainly not clear that China's model is what the U.S. wants to emulate. In fact, it's probably a non-starter. For example, just like other areas of tech, China restricts AI that has the potential to undermine social order, which in China means a critique of the government in any way. But the point, which is true, is that China has been dealing and thinking about these issues for somewhat longer than other global players around the world.

12:26Now, one additional interesting dimension of the AI battle vis-a-vis China and the U.S. is that the larger policies around silicon chips are impacting this as well. This is a recent Wall Street Journal article called U.S. sanctions drive Chinese firms to advance AI without latest chips. Research in China on workarounds, such as using software to leverage less powerful chips, is accelerating. Over the last 20 years, chip manufacturing moved almost entirely to places far away from the U.S., including most notably Taiwan, where a huge percentage of the world's global chips are fabricated. A couple of years ago, the U.S.

13:02started to realize that it was potentially a really dumb geostrategic move to outsource such a critical technology to a place not only far away, but in a highly contentious part of the world. Since then, the Biden administration has been putting serious effort into moving chip manufacturing back on shore, offering hundreds of billions of dollars in incentives and tax breaks to do so. This WSJ piece argues that those policies are having an impact when it comes to China's AI industry. From the article, quote, For Chinese companies, the issue is more critical. U.S. sanctions have cut them off from the most advanced chips made by the likes of NVIDIA, and they have rapidly consumed existing American chip stocks to create their own chat GPT equivalents.

13:40This Forbes piece argues that, quote, the geopolitics of AI chips will define the future of AI, and I don't think that's entirely hyperbole. Now, on top of that, there are also potential other restrictions that the U.S. is considering when it comes to Chinese AI. This piece published just today in Fox Business says U.S. weighs restrictions on investment in Chinese AI firms. U.S. firms' investment in Chinese AI companies could face greater scrutiny under screening regulation that may be released soon. This is, again, both part of a larger trend in which the U.S. government is taking a much stronger look at how American business is entangled with Chinese business, but is also unique and distinct to the AI space, which is seen as an extremely geostrategically important area.

14:19In Europe so far, a lot of the discourse has been around how ChatGPT and other generative AI platforms can comply with existing rules like GDPR. For example, right now the European Parliament is working on something they're calling the AI Act, which they say is trying to, quote, steer the development of very powerful artificial intelligence in a direction that is human-centric, safe, and trustworthy. A note reviewed by the Wall Street Journal says, with the rapid evolution of powerful AI, we see the need for significant political attention. And as they said about their recent crypto legislation, MECA, the EU is hoping that their bill could, quote, serve as a blueprint for other regulatory initiatives in different regulatory traditions and environments around the world.

14:58Italy, of course, notably recently won concessions from OpenAI around chat GPT, including adding new age verification and other sort of privacy-preserving measures. A recent G7 meeting also spent a significant amount of time focused on AI. And I think that the biggest takeaway from this is not any one specific policy that seems like it might be on the horizon, but the fact that this is an issue that they see as necessary to coordinate on, that it's not just a nation-by-nation issue, or at least these governments are trying to make it an international issue. Now, when it comes to other parts of the world, there's a real concern that AI advances could expand the gap between the US, China, and everyone else.

15:35A recent conference in Kigali, Rwanda had more than 2 ,000 researchers and engineers gather to discuss AI, and this was one of the key themes. AI industry leader Yoshua Bengio said, it's become obvious that in order to bring the potential benefits of AI to everyone, we need everyone to be a part of it. And then of course, there's the US, which is seeing increasing attention focused on AI. Last week, the president and vice president Kamala Harris met with AI CEOs, and there have been plenty of critiques of who was invited and who wasn't and what they discussed and what they didn't to go around.

16:05But the point is that when you see this sort of high-level meeting, it suggests that the White House is taking it seriously. So to try to sum up, in many ways, the geopolitics of AI are simply following other geopolitical realignments, the move from a unipolar to a multipolar world, countries jockeying for position within that new framework, regulatory coordination, and very different regulatory approaches in different countries around the world. All of those things, AI is kind of coming into an existing framework. However, AI also presents radical new challenges. It is an issue that crosses different spheres.

16:39It is not just a technology issue. It is an economic issue. It is a productivity issue. It is an equality issue. And you better believe that these countries know that it's a national security issue as well. It is, in fact, big enough as an issue that I believe that the existing paradigms of geopolitics might be challenged in some ways by it. Although whether that means countries fracture even more in a vicious race to get ahead, or they come together to try to actually create some guardrails around it remains to be seen. But either way, in the same way that the last six months has seen a massive increase in consumer awareness of AI as they start to use these tools, I think that the next six months is going to see a massive increase in government attention on this incredibly important issue.

17:18All right guys, hopefully that gives you a baseline of the geopolitics of AI as they stand today, this will be, I anticipate, a fast-moving space. If you're enjoying the AI breakdown, please like this video, subscribe to the channel, go listen to the podcast, or subscribe to the newsletter. Any way that you want to consume it is cool with us. Until next time, peace.

17:52Meet today a video of the даль lenghtar was made up and bath semblance on the streets with frictionと HAVE R currency on their web browser BECOMEN SO IF YOU PidiON THANK YOU

18:17THANK YOU Thank you.