The AI Daily Brief: Artificial Intelligence News and Analysis - The Geopolitics of Artificial Intelligence

Episode Date: August 31, 2023

On today's episode, NLW looks at news that the US has pushed Nvidia and AMD to stop AI chip exports to certain Middle Eastern countries as part of their attempt to deny China access to compute. Before that on the Brief: can AI save Silicon Valley from the post-ZIRP era? Today's Sponsor: Supermanage - AI for 1-on-1's - https://supermanage.ai/breakdown ABOUT THE AI BREAKDOWN The AI Breakdown helps you understand the most important news and discussions in AI.  Subscribe to The AI Breakdown newsletter: https://theaibreakdown.beehiiv.com/subscribe Subscribe to The AI Breakdown on YouTube: https://www.youtube.com/@TheAIBreakdown Join the community: bit.ly/aibreakdown Learn more: http://breakdown.network/

Today on the AI breakdown, we're looking at the geopolitics of AI as the U.S. blocks exports of AI chips to the Middle East. Before that on the brief, a new AI unicorn tells us about the state of the AI startup market. The AI breakdown is a daily podcast and video about the most important news and discussions in AI. Go to Breakdown.network for more information about our YouTube, our newsletter, and our Discord. Welcome back to the AI breakdown brief, all the AI headline news you need in around five minutes. We kick off today with a tweet from investor Yaron Samid who writes, The Reports of the Death of Unicorns are greatly exaggerated. Hats off to the team at AI21 Labs.

Now, the story that Yaron is referencing is that the generative AI startup AI 21 has raised $155 million Series C at a $1.4 billion valuation, a nice newly minted AI unicorn. The story is interesting in one part because it's a big fat number, but more than that because it actually has a little bit to tell us about the state of the AI startup space. One of the things that's been interesting to watch this year is how VC and startup thinking around AI has been challenged in practice. Many of the companies that raised money in the immediate wake of the release of ChatGPT have effectively ended up being little more than a wrapper around the OpenAI API API and haven't necessarily done well. We've spent a bunch of time on other episodes talking about why that is, ranging from the availability of open source models that enterprises themselves have used to customize to competition from the big players. See, for example, ChatGPT Enterprise release this week.

But by and large, there has been a sense that perhaps, despite AI seeming like the last great hope for the Silicon Valley funding model that proliferated during the zero interest rate era, maybe even AI couldn't actually sustain that shift. Indeed, on August 30th, Wednesday this week, the Wall Street Journal wrote a piece called AI startup buzz is facing a reality check. Venture investors are realizing that generative artificial intelligence might not be enough to stem years-long startup downturn. Now, there are two things at work here. One is AI startups how hyped they are, how immediately successful they've been, how much they will sustain

portfolio returns. But then there is the much larger thing going on, at least as relates to the venture capital industry, which is, of course, this fundamental secular shift in the environment in which venture capital is operating. In the decade plus that followed the global financial crisis, the zero or near zero interest rate policies of the Federal Reserve and other governments around the world created a scenario in which more and more capital flowed into the venture capital asset class because companies and investors were forced to move farther out on the risk spectrum to achieve the yield that they needed. The availability of more total venture capital dollars meant one that startups were able to raise more money at higher valuations and that,

Two, the dynamics of the industry were such that because of that proliferation and availability of capital, growth was prioritized over profitability. Think about it this way. If you're a company who knows there's another big dose of venture capital just around the corner, as long as you hit certain growth metrics, well, then you do whatever it takes to hit those growth metrics, even if it means lighting money on fire. If, on the other hand, you've just raised the last amount of capital that feels available to you, you're naturally going to prioritize things like revenue and profitability a hell of a lot more. Broadly speaking across the venture capital sector,

we've seen a shift in response to the fastest rate hiking cycle in 40 years of companies no longer thinking in terms of growth at any cost mindsets, and instead once again starting to prioritize things like revenue and profitability. Now, the one holdout to this model seemed to be artificial intelligence, where VCs and investors basically couldn't put enough money fast enough into any startup that slapped AI on its pitch deck. However, what the industry is finding is that even with, the hype and excitement around AI, it's simply not enough to fight against those larger forces.

Bringing it back to AI21 then, what makes it interesting is that it shows a number of differences as relates to many of the startup funding announcements that you were seeing maybe a few months ago. As TechCrunch describes, AI21 Labs flagship project is AI21 Studio, a pay-as-you-go developer platform for building custom text-based business apps off of AI 21's proprietary text-generating AI models, including its cutting-edge Jurassic 2 model. Now, two things that are notable about that. One, pay-as-you-go developer platform, i.e. there is already a revenue model. And two, this is based on proprietary AI models, not just something that's a rewrapper of OpenAI. The reason that I think it's worth spending a little bit more time on this than usual for the brief section of this show is that as you heard with that Wall Street Journal piece, there is this strong narrative in mainstream media right now, that the AI hype has totally died down, that it's underperforming, yada, yada, yada, yada. Now, I've talked extensively about why I think parts of that are true, but I think that the narrative is getting way out of hand. and way over-exaggerated in media right now.

And it would be an easy mistake to make to over-correct from the perhaps over-exuberance of a few months ago. Just by way of example, if you look at this Wall Street Journal piece, it's the same example cited every single time in every article trying to make this point about hype dying down, which is really just chat GPT usage going down in June, and Jasper and one other AI startup having some layoffs. I don't know about you guys,

but those data points alone are not enough to convince me that the entire hype around an entire industry hasn't paid off, and that we should just go pay attention to something else. Now, one dimension of the AI space that's interesting is that it's not just traditional tech companies that are a big part of the community. Recognizing that, Andrewson Horowitz has announced a number of open source grants. In a blog post released yesterday, they write, we believe artificial intelligence has the power to save the world, and that a thriving open source ecosystem is essential to building this future. Thankfully, the open source ecosystem is starting to develop and we are now seeing open source

models that rival close source alternatives. Hundreds of small teams and individuals are also working to make these models more useful, accessible, and performant. However, the people behind these projects often don't have the resources available to pursue their work to conclusion or maintain it in the long run. The situation is more acute in AI than traditional infrastructure, since even fine-tuning requires significant GPU computing resources, especially as open-source models get larger. Given that, they announced the A-16 Open Source AI grant program and announced the first batch of recipients, which are heavily emphasized around fine-tuning LLMs and foundation models for various purposes, including to run locally or to explore things like synthetic data.

Now, speaking of shifting media narratives, alongside the broader AI hype is fading stories, you're also seeing more stories featured in big news outlets around companies walking away from AI content experiments, as opposed to a couple months ago, where the stories were all about companies trying AI content experiments. The latest comes this morning from CNN, Gannett to pause AI experiments after botched high school sports articles. CNN writes, Newspaper Chain Gannett has paused the use of an artificial intelligence tool to write

high school sports dispatches after the technology made several major flubs and articles in at least one of its papers. Several high school sports reports written by an AI service called Lead AI and published by the Columbus dispatch earlier this month went viral on social media this week and not in a good way. One notable example preserved by the Internet Archive's Way Back Machine began. The Worthington Christian bracket bracket bracket winning team mascot bracket bracket defeated the Westerville North, bracket bracket bracket bracket losing team mascot bracket two to one in an Ohio boy soccer game on Saturday. Says CNN, the reports were mocked on social media for being repetitive, lacking key details using odd language, and generally sounding like they'd been written by a computer

with no actual knowledge of sports. Now, here's what I have to say about this. I have no dog in the fight about how AI should be used or shouldn't be used when it comes to media. I think certainly the upside of having better coverage of more niche things, such as high school sports or just local issues in general, could be one of the ways that AI actually really adds value. In other words, this is not replacing a beat that a reporter already has. It's reopening a beat that has been replaced over the last 20 years as we've moved towards more general national rather than local news. But holding aside even that, when I look at this article that is being mocked, that has bracket bracket winning team mascot and bracket bracket losing team mascot in the first line,

it feels to me as though the publisher in this case forgot that a news story isn't just the product of a writer, but also the product of an editor and a fact checker. Sure, the AI may not be as sophisticated as one would like if it would make that mistake in the first place, but someone either published a thing or made the stupid decision to not check it before it was published. In either case, I would put the blame squarely with the editorial team over the artificial intelligence platform. But like I said, I don't have a dog in this fight. I just think it's important to not be stupid about what's stupid about AI, given that there are some real honest-to-god battles to be had. Lastly, today, because it's getting some buzz, an AI-powered drone has beaten,

a human champion of a drone racing league. NPR's All Things Considered, today researchers in Switzerland unveiled a small drone powered by artificial intelligence that can outfly some of the best human competitors in the world. A quadcopter drone equipped with an AI brain whipped its way around an indoor racecourse in a matter of seconds. In 15 out of 25 races, it was able to beat its human rival, according to research published today in the journal Nature. Says Elia Kaufman, an autonomy engineer at Skydeo, a drone company based out of California who worked on the drone while at the University of Zurich, quote, this is the first time that an AI has challenged and beaten human champions in a real

world competitive sport. Now, there is something notable about this. I don't want to deny that. Obviously, it has implications for other areas of AI and robotics. Think, for example, self-driving cars. I do have to laugh a little bit, however, because I've watched for a decade, promoters of drone racing leagues and professional drone sports, desperately try to turn this into, as Elliott calls it, a real world competitive sport, with what I would characterize as very limited success. But like I said, still significant. Certainly you have to think that militaries in places like China and the U.S. are watching this study closely. In any case, that is going to do it for today's AIA breakdown brief. If you enjoyed this, do me a favor. If you haven't subscribed to the

podcast yet, please go do it. You can find it on iTunes. You can find it on Spotify, basically anywhere you listen to podcasts. And once you've subscribed, if you want to throw a five-star rating my way, I would so appreciate it. It makes a big difference when new people are trying to decide if they are going to check out the show. In either case, thanks for listening or watching, and I'll be back soon with the main AI breakdown. Before we get into the main AI breakdown, I want to tell you about today's sponsor, Supermanage. If you work in a professional setting, you probably have some version of a one-on-one meeting, either with the people that work for you or the people that you work with.

Unfortunately, all too often, those one-on-one meetings become glorified catch-up calls. Don't you wish you could jump right to the stuff that really matters? that's where SuperManage comes in. Supermanage AI magically distills your team's public Slack channels into a real-time brief on any employee, any time. Catch up on contributions, work in progress, challenges they're facing, sentiment, everything you need to show up ready for a truly meaningful conversation, and it's completely free.

Visit supermanage.a.i forward slash breakdown today to start making the most of your one-on-ones. And thanks again to Supermanage for sponsoring the AI breakdown. Welcome back to the AI breakdown. Today we are talking about what I think is one of the most interesting aspects of artificial intelligence, which is the way in which it's impacting geopolitics and global political relationships. Now, the specific genesis of our conversation today is news from Nvidia that was repeated by AMD that the U.S. government came to them in the second quarter and asked them to stop selling their chips, specifically their A-100 and their H-100 to a number of countries in the Middle East. Now, the countries in the Middle East weren't specifically named, but it appears that this is not about denying AI compute to the Middle East specifically, but instead, part of the larger strategy to deny advanced

chips to China. We'll get a little bit more into the specifics of this news story in just a moment, but first, let's put this in a larger context. In December 2022, just after ChatchapT was released, the Brookings Institution released a paper called the Geopolitics of AI and the rise of digital sovereignty. The paper threw it all the way far back to September 2021, when the Trump, Trade and Technology Council of the U.S. and the EU held their first summit. Now, obviously, tensions with China had been ratcheting up for a number of years at that point. Remember, one of the things that was notable about the 2020 elections was that one of the only things that Biden and Trump agreed upon was that the U.S. needed to take a strong hand against China. In fact, one of the points of

competition was not a difference of opinion there, but who would or wouldn't be more opposed to the country? Now, bringing it back to this Trade and Technology Council meeting in 2021, following that meeting, the U.S. and the EU, quote, declared their opposition to artificial intelligence that does not respect human rights and reference rights infringing systems, such as social scoring systems. During the meeting, they clarified that, quote, the United States and European Union have significant concerns that authoritarian governments are piloting social scoring systems with an aim to implement social control at scale. These systems pose threats to fundamental freedoms and the rule of law, including through silencing speech, punishing peaceful assembly and other expressive activities, and reinforcing

arbitrary or unlawful surveillance systems. Now, as Brookings makes clear, the target of the criticism was clearly China. And what was happening was that a new technology, or at least a technology that was coming into greater prominence in the public conversation, artificial intelligence, had real implications for how much control China in this case, but really any authoritarian government had in the ability to surveil its citizens. AI is, of course, a technology that is predicated upon and takes advantage of huge, huge amounts of data. China is no stranger to that data, as the U.S. and the EU pointed out, using it as part of a larger social credit scoring system, and so the concern was that AI, harness to these authoritarian ends, could further undermine human rights. Now, this Brookings' research

paper in 2022 basically argued that these increasing geopolitical concerns, quote, threatened existing forms of interconnectivity, causing markets for high technology to fragment, and to varying degrees retrench back into the nation-state. And in many ways anticipating what the conversation has been this year, the piece writes, the ideological differences between the three great powers could have broader geopolitical consequences for managing AI and information technology in the years to come. Control over strategic resources such as data, software, and hardware has become paramount to decision makers in the United States, the European Union, and China, resulting in a neomercantilist-like approach to governance of the digital space. Resurfacing neomercantilus ideals are

most visible in the ways that trade and semiconductors is being curtailed, but they are also apparent in discussions over international data transfers, resources linked to cloud computing, the use of open source software, and so on. These developments seem to increase fragmentation, mistrust, and geopolitical competition. Now, Brookings predicts, quote, as technological decoupling deepens, China will seek to maintain its goal of achieving self-sufficiency and technical independence, especially from high-tech products originating in the U.S. The piece concludes, undoubtedly AI will continue to revolutionize society in the coming decades. However, it remains uncertain whether the world's countries can agree on how technology should be implemented for the greatest

possible social benefit. As stronger forms of AI continue to emerge across a wider range of use cases, securing AI alignment at the international level could be one of the most significant challenges of the 21st century. Over the course of 2023, we've seen many of the assessments of Brookings in that paper from the end of last year come true. The U.S. has put increased pressure to deny China access to AI chips, and broadly speaking, this is increasingly recognized as an issue of geopolitical importance. On August 7th, Goldman Sachs wrote a post called AI in the crosshairs, how geopolitics shapes technology. Somewhat stating the obvious, they write, artificial intelligence is viewed by some political leaders as a technology that potentially carries greater geopolitical

significance than previous technological shifts. They continue, earlier transitions like the adoption of cloud computing, expansion of social media, and the proliferation of smartphones, rose initially with minimal governmental interference. However, as they point out, the White House has acknowledged that preserving the U.S.'s lead in the science and technology of AI is both a domestic and a national security issue. GS again points out, given the strategic applications of AI, this technology is now a key factor in the ongoing competition between the United States and China. They write, with AI leadership a clear priority for both the U.S. and China, critical commodities for AI may continue to be subject to export restrictions. AI will likely receive similar treatment to other

dual-use technologies like nuclear and aerospace. From the emergence of an escalating high-tech trade competition to government subsidization driving the boom in Chinese semiconductor startups and American chip manufacturing, state involvement has already begun to reshape global markets. Now, one of the interesting things that this article points out is that in addition to just AI chip restrictions, one of the challenges for China when it comes to AI and specifically LLMs is the difficulty in controlling them and shaping their messaging to conform with the CCP party line. As GS writes, LLMs depends on. on vast volumes of text for development and constraints on permissible content could compromise

the performance of Chinese LLMs compared to Western LLMs that have ready access to the full open web. Furthermore, outputs of LLMs are unpredictable and difficult to control. The Chinese government's focus is evident in a recent regulatory framework, which aims to impose disclosure requirements, liability for output and restrictions on data used in the development of LLMs. Now, this is interesting in context of another piece of news from today. Reuters and others are reporting that five Chinese tech firms, including Baidu and Cense time group have been allowed as of Thursday to finally launch their AI chatbots to the public. Now, this was obviously Herald in Chinese public markets, where shares of both Bidu and Centsestime

jumped a couple of points. And apparently on top of these five firms which released something, Chinese media was reporting that a total of 11 firms had received approvals from the government. Notably, that includes TikTok owner Bight Dance as well as 10 cent holdings, but the big one that's not clear is whether Alibaba has received approvals. A financial times op-ed today nailed what's going on in China, writing, fear of falling behind forces sensors to tolerate chatbots. This is obviously going to be one of the really interesting dimensions within China of just how much control they're able to assert over an inherently much more difficult to control type of technology. Now, this gets us back to the other piece of news around the U.S. restricting

NVIDIA and AMD chips to the Middle East. A statement from NVIDIA read, during the second quarter of fiscal year 2024, the U.S. government informed us of an additional licensing requirement for a subset of A-100 and H-100 products destined to certain customers in other regions, including some countries in the Middle East. And as many of the articles reporting this news have pointed out, it's really about denying China the ability to get chips through allied intermediaries that have a relationship with both the U.S. and specifically U.S. companies like NVIDIA and AMD as well as with China. Now, Nvidia for its part has warned about some of these restrictions. The cynical view is, of course, that it hits their bottom line, which it absolutely does. But the argument, for example, from Jensen Huang,

the CEO of NVIDIA is that China will just figure out alternatives, so in some ways the United States may be cutting itself off from a type of control. Huang said, if China can't buy from the U.S., they'll just build it themselves. Now, still, it's pretty clear that his was an economic argument more than a political argument. He continued the U.S. has to be careful. China is a very important market for the technology industry. If we are deprived of the Chinese market, we don't have a contingency for that. There is no other China. There is only one China. He said if the American tech industry requires one-third less capacity due to the loss of the Chinese market, no is going to need American fabs. We will be swimming in fabs. If they're not thoughtful on regulations,

they will hurt the tech industry. Compare this with an article from yesterday from the South China Morning Post called Tech War, China exploring ways to make its own AI memory chips despite U.S. sanctions. The article reads, China is exploring ways to produce its own high bandwidth memory, the next generation of memory chips tailored for AI processors, as it pushes ahead with a semiconductor self-sufficiency drive amid U.S. sanctions. While it will be an uphill battle to catch up with global leaders, The Chinese government has determined that the country must become self-sufficient in HBMs, even though it may take years. Now, interestingly, it is not just China in the U.S. who are thinking about the geopolitical role of AI and chips. Reuters from a week ago writes Germany plans to double AI funding and race with China and U.S.

The article reads, Germany plans to almost double its public funding for artificial intelligence research to nearly a billion euros over the next two years, as it attempts to close a skills gap with sector leaders China and the United States. Germany envisages creating 150 new university labs for AI research, expanding data centers, and making accessible the kind of complex public data sets from which AI techniques can tease out new insights, a major undertaking in a country where cash transactions are common and the facts is not yet extinct. Still from a scale perspective, Reuters points out, quote, it is dwarfed by private spending in the U.S., which reached 47.4 billion in 2022, almost double Europe's total spend and well ahead of China's 13.4 billion.

Meanwhile, as I pointed out in previous episodes, in addition to the domestic and economic competitive aspects of the AI battle, the military side continues ahead full steam as well. On August 19th, the Wall Street Journal wrote a piece, for example, about the race between China and the U.S. around AI-powered drone swarms. The takeaway of all this is just a reminder that AI, more than most technologies than we've seen, and certainly in the last few decades, has immediate, clear, and profound geopolitical implications. This adds complications to every demand. dimension of the industry, from who and how it gets funded, to policy and regulations, to the business operations of startups and companies within it, et cetera, et cetera, et cetera. It's a really interesting space, and I will, of course, continue to keep you updated as shifts at this intersection of AI and geopolitics start to occur. For now, that is going to do it for today's episode. Thanks for listening or watching as always. And until next time, peace.