The AI Daily Brief: Artificial Intelligence News and Analysis - America's AI Action Plan

Starting point is 00:00:00 Today on the AI Daily Brief, America's AI Action Plan. Before that in the headlines, AI Search is growing super, super fast. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. Hello, friends, quick announcements today. First of all, thank you to today's sponsors, Blitzy, Plum, Vanta, and Superintelligent. To get an ad-free version of the show, go to patreon.com slash AI Daily Brief. And if you are interested in sponsoring the show, shoot me a note at NLW at Breakdown. network. Welcome back to the AI Daily Brief Headlines edition, all the daily AI news you need in

Starting point is 00:00:37 around five minutes. Today we kick off with the not all that surprising news that AI search is growing incredibly fast. According to market intelligence firm Dato's 5.6% of U.S. search traffic flowed through AI search providers last month. Now that is, of course, still a small proportion compared to traditional search, but market share has doubled compared to a year ago, and the numbers are even starker when you just look at early adopters. Dados began tracking a separate cohort of people who are using LLMs in their desktop browsers in April of last year. Visits to LLMs are now 40% of that cohort's browser usage up from around 24% a year ago. And these figures also don't take into consideration searches that are satisfied with Google's

Starting point is 00:01:18 AI overviews as those go through the normal search front end. The very clear point is that while we're not there yet, the growth trajectory absolutely suggests that LLMs will become the dominant form of search within a few years' time. Eli Goodman, the CEO of Dados, believes this is a sea change for the web comparable to Google's launch and the emergence of the first social media platforms. It's not just, in other words, that people are switching to a different search engine. It's that they're rethinking in a fundamental way how they engage with the Internet. Goodman said, over 90% of all the AI searches are what we call informational or productivity-based.

Starting point is 00:01:52 Help me solve this problem. Help me answer this question. In contrast, the majority of traditional searches are about finding a particular destination on the web. Now, one of the groups that this has big implications for are, of course, the marketers. There is a ton of discussion in the SEO world about how to switch over to generative engine optimization, but there's clearly a sea change afoot. Some optimistically think that AI search will be additive rather than replacing traditional search. Andrew Lipsman, the founder of consulting firm Media, ads, and commerce, thinks that this will be close to the mobile boom.

Starting point is 00:02:21 Many feared consumers would ditch the desktop and go mobile only. Instead, what happened was that desktop internet traffic stayed relatively flat while mobile traffic surged. So is it possible that that's what will happen with AI search, where because it serves a different purpose than traditional search, especially through the use of tools like deep research, it won't be a one-to-one replacement, it'll just create a new pillar of how we interact. Obviously, one of the biggest sources of those new searches is ChatGBT, BT, where users now send over 2.5 billion prompts every single day. That's coming up on a trillion requests every year. OpenAI shared these figures with Axios, and while Google doesn't share their stats on a daily level,

Starting point is 00:02:59 they did recently reveal that they're currently receiving around 5 trillion queries a year, which puts OpenAI at creeping up on a fifth of Google searches. What's more, it appears that usage is growing at a very significant pace given that in December, Sam Altman said that users send over a billion queries a day. I'm no mathematician, but going from 1 to 2.5 billion, means that usage is more than doubled in the last seven months. The numbers were reported as part of Altman's trip to Washington this week. He's scheduled to speak at a Federal Reserve event on Tuesday to discuss how,

Starting point is 00:03:29 AI is impacting jobs, where sources close to Altman said that his view is, quote, not all dummerism downside and not hand-waving away concerns, but focused on democratizing benefits, not concentrating them in the hands of a few. Today, on Wednesday at the time of recording, Altman will be present at the White House for a keynote from the president at the AI Race Summit. Coinciding with that is the AI action plan, which will be the subject of today's main episode. A source told Axios, over the course of the week, Open AI will make the case that democratizing the economic benefits begins with putting these tools in the hands of people to do stuff, so they have an opportunity to participate in the prosperity.

Starting point is 00:04:04 Next up in the headlines, an update on the talent wars. Earlier this week, of course, we discussed OpenAI and Google putting up gold medal performances in the International Math Olympiad, and brutally it turns out that three Google researchers who designed the model were being poached by meta. The information reports that the trio worked on a version of Google's Gemini model that leveraged advancements in reasoning to bring home the gold. They reported the moves as meta's latest hires, so it's entirely possible given how that that man operates that Zuck could have been making calls as soon as the news broke.

Starting point is 00:04:34 Separately, CNBC reports that Microsoft is rating Google DeepMind as well. Sources said that Microsoft had hired around two dozen researchers from DeepMind over recent months. On Tuesday, a 16-year Google vet, Amr Sabramanya, who was most recently VP of Engineering for Gemini, announced on LinkedIn that he had joined Microsoft as a corporate VP. The reporting suggests that Microsoft AI CEO Mustafa Sullyman, who was, of course, himself a co-founder of Deep Mind, is more aggressively staffing up his AI division to get back in the race. One more big company update, Amazon, has acquired an AI wearable startup called B. The product is a watch that listens to your conversations and can generate summaries or transcripts,

Starting point is 00:05:13 similar to a number of the different AI pins and pendants. The big selling point is a $50 price tag, which comes in below the competition. The transaction apparently hasn't closed yet, and Amazon didn't discuss. close the size of the deal, but their spokesperson did mention that B employees have received offers to join Amazon as part of the deal. And indeed, that could be the rationale here, that this is a hiring play rather than Amazon wanting to launch a budget AI wearable. As you well know, Amazon has been struggling to get Alexa up to any sort of impressive or even frankly usable standard, so it doesn't seem like bringing in additional talent with experience in Ambient AI is the worst

Starting point is 00:05:46 idea. Now, later today, in the main episode, we're going to talk a lot about AI and geopolitics. And one interesting sub-story and all that from earlier this week is a leaked memo from Anthropic CEO Dario Amaday, where Dario is basically shifting Anthropics policy embracing his company to start taking investment from organizations in the Gulf states. Wired received a slack message sent to Anthropic staff on Sunday, in which Amade weighs up the pros and cons of accepting money from the Middle East. He wrote, among other things that it would likely enrich quote-unquote dictators, which he said was a, quote, real downside and I'm not thrilled about it. However, he concluded that Anthropic would take the money, saying, Unfortunately, I think no bad person should ever benefit from our success is a pretty difficult principle to run a business on.

Starting point is 00:06:29 This was a multi-page memo that sort of seemed to read like a post on Less Wrong, carefully weighing up the pros and cons across any number of different variables, but ultimately it just came down to the resource needs for the next leg of the AI race. Amadei wrote, There is a truly giant amount of capital in the Middle East, easily $100 billion or more. If we want to stay on the frontier, we gain a very large benefit from having access to this capital. Without it, it's substantially harder to stay on the frontier. In a section titled Erosion of Standards, Amade said that the reason that they had previously,

Starting point is 00:06:58 quote, vociferously pushed for not allowing big data centers in the Middle East, was because, quote, without a central authority blocking them, there's a race to the bottom where companies gain a lot of advantage by getting deeper and deeper in bed with the Middle East. Dario seemed to have a deep reluctance, but also realized the company's principles were holding it back, commenting, Unfortunately, having failed to prevent that dynamic at a collective level, we're now stuck with it as an individual company, and the median position across the other companies

Starting point is 00:07:22 appears to be outsourcing our largest 5-gagawatt training runs to UAE and Saudi is fine. That puts us at a significant disadvantage, and we need to look for ways to make up some of that disadvantage while remaining less objectionable. I really wish we weren't in this position, but we are. In a section titled Com's headache, which is exactly what this whole leaked memo was giving Anthropics' comms teams, Amade argued that the media, Twitter, and the outside world is, quote, always looking for hypocrisy, while also being very stupid and therefore having a poor understanding of substantive issues. I have a lot of thoughts about this. I think that there are many reasonable positions to have when it comes to this particular question. I think that anyone who says

Starting point is 00:07:59 that dealing with certain regimes in the Gulf states is completely obviously morally clear, probably hasn't been paying a ton of attention. At the same time, I think it's a lot more complex than lumping them all together, painting everything with a broad brush, and calling everyone who doesn't get it stupid. The Gulf has a very particular and interesting geostrategic role in the world that we're headed into. It sits both metaphorically and literally in between the U.S. and China. The implications for where it decides to put its emphasis and capital go way beyond just the AI race and will shape the next couple of decades in many more ways. It's deserving of an entire show, which maybe I will do at some point, but for now, you've got to think that this whole

Starting point is 00:08:38 discourse is going to make it harder, not easier for Anthropic to operate in this very important part of the world. Certainly the White House was plenty happy to jump on it with AI czar David Sacks writing, Anthropic criticized President Trump's AI investment deals with the Middle East, but now it seeks investment from those very same countries. Perhaps it now realizes the absurdity of refusing to do business with the Gulf states. All we will do is push them into the arms of China. Now with that interesting ending point, let's dive into what is a very interesting and packed main episode. This episode is brought to you by Blitzy. If you're a technology leader, here's something that probably sounds familiar. Your organization's competitive edge is buried in

Starting point is 00:09:15 legacy code that desperately needs modernization, but the resources required feel out of reach. That was the case for a global investment analysis firm. They needed to migrate 70,000 lines of complex MATLAB financial algorithms to Python. Algorithms that drive investment decisions for trillions in assets. Their estimate? Months of high-cost specialized engineering work. Instead, they partnered with Blitzy. Blitzy's autonomous AI preserved mathematical precision and generated over 80, percent of the new codebase, completing the migration with just five days of engineering time. They cut the timeline by 95% and saved 880 engineering hours. If your organization is facing similar modernization challenges, visit blitzie.com to schedule a consultation and discover how AI

Starting point is 00:09:56 power development can transform your technical capabilities. Today's episode is brought to you by Plum. You put in the hours, testing the prompts, refining JSON, and wrangling nodes on the canvas. Now, it's time to get paid for it. Plum is the only platform designed for technical creators who want to productize their AI workflows. With Plum, you can build, share, and monetize your flows without giving away your prompts or configuration. When you're ready to make improvements, you can push updates to your subscribers with a single click. Launch your first paid workflow at useplum.com. That's Plum with a B and start scaling your impact.

Starting point is 00:10:33 Today's episode is brought to you by Vanta. In today's business landscape, businesses can't just claim security, have to prove it. Achieving compliance with a framework like SOC 2, ISO-27-01, HIPAA, GDPR, and more, is how businesses can demonstrate strong security practices. The problem is that navigating security and compliance is time-consuming and complicated. It can take months of work and use up valuable time and resources. Vanta makes it easy and faster by automating compliance across 35-plus frameworks. It gets you audit-ready in weeks instead of months and saves you up to 85% of associated costs. In fact, a recent IDC White Paper found that Vanta customers achieved $535,000 per year in benefits,

Starting point is 00:11:13 and the platform pays for itself in just three months. The proof is in the numbers. More than 10,000 global companies trust Vanta. For a limited time, listeners get $1,000 off at vanta.com slash NLW. That's VANTA.com slash NLW for $1,000 off. Today's episode is brought to you by Superintelligence, specifically Agent Readiness Audits. everyone is trying to figure out what agent use cases are going to be most impactful for their business, and the agent readiness audit is the fastest and best way to do that.

Starting point is 00:11:44 We use voice agents to interview your leadership and team, and process all of that information to provide an agent readiness score, a set of insights around that score, and a set of highly actionable recommendations on both organizational gaps and high-value agent use cases that you should pursue. Once you've figured out the right use cases, you can use our marketplace to find the right vendors and partners, and what it all adds up to is a faster, better agent strategy.

Starting point is 00:12:08 Check it out at B-Super.AI or email agents at B-Super.A.I to learn more. Welcome back to the AI Daily Brief. Today I was going to do the main episode all about this new Anthropic report about building AI in America. That report, as we'll talk about, gets into energy concerns, geopolitical competition, and is a really interesting document that has even more interesting context based on some leaked memos of Anthropics earlier in the week. But today, as I was preparing to record, the White House dropped their own 28-page report on America's AI Action Plan, which of course is no coincidence. It's clear that Anthropic was timing their report alongside this. But what it sums up as is an interesting

Starting point is 00:12:50 set of documents that show how both the U.S. government and the big labs are thinking about the key geostrategic challenges and just in general the infrastructure challenges that face the AI industry as it moves to the next levels. There is a lot of commonality despite different sources, and so that's what we're going to talk about today. And let's actually start with the Anthropic Report. Here's the nut of it. For the United States to lead the world in AI, they write, we must make substantial investments in computing power and electricity that make it possible to build AI in America. As technology evolves, the required resources, infrastructure, and regulatory conditions need to change too. Now, right from the beginning, what you get here is that when they say

Starting point is 00:13:30 build AI in America, they are not using the word build metaphorically. They are talking about literally the physical infrastructure required to produce AI here. Right at the top of the list, their first focus is the energy requirements for global AI leadership. Anthropics big point here is not only that new energy generation in the U.S. is lagging. It is getting absolutely smoked by China. While the U.S. is lagging and bringing energy generation online, China is rapidly building energy infrastructure for AI, having added over 400 gigawatts of power capacity last year, compared to the just several dozen gigawatts added in the U.S. Now, a question that might come up, since neither I nor most of you are energy experts,

Starting point is 00:14:12 is what sort of power are we actually talking about? Obviously, 400 versus several dozen makes clear the gap, but what are the requirements going to be? Well, Anthropics says that they project they're going to need 2 gigawatt and 5 gigawatt data centers to develop a single advanced model for just themselves, in 27 and 28. They expect that by 2028 in total, the frontier AI labs are going to require 20 to 25 gigawatts just for training demands. The comparison they use is New York City's peak electricity demand with this being twice that. What's more, they say that that's just for training

Starting point is 00:14:44 and at least as much will be needed for inference. In other words, the compute that's used while people are actually using these models. In total, then, the USAI sector is going to need at least 50 gigawatts of electric capacity by 2028. Josh Kelly writes, that's about one Netherlands Netherlands, Sweden, Argentina, or Taiwan. Anthropic, on its own, will need a New Zealand. Now, it's not so much that America doesn't have the power to keep pace or the ability to bring more power online. It's that China is spinning up new power supply at such a rapid pace that they're completely unconstrained. They can spin up new industrial plants, AI, and domestic electrification all at the same time without taxing supply. The U.S. at this point isn't really close to AI-induced

Starting point is 00:15:22 brownouts, but at the pace we're on, it's starting to become a possibility. Anthropic are therefore calling for an all of the above approach to adding more power generation. Rather than focusing on just green tech or just natural gas or just nuclear, Anthropic wants to see everything expand much faster. Now, they do note that AI is the perfect catalyst for pushing emerging energy technology like geothermal and advanced small nuclear, but they really want all of it. Anthropic also calls for, in this report, streamlined permitting, to help data centers be built fast. They have a long section about the challenges of building energy in the U.S., including the six main types of pre-construction permits, transmission approvals, and more.

Starting point is 00:16:01 In their report, Anthropic proposes two big pillars of American AI infrastructure. The first is to build large-scale AI training infrastructure. For that, they want to see federal lands be made available as options for AI infrastructure construction to avoid the years-long process of state and local zoning. They also want to see public-private partnerships for things like expedited power line buildouts. Their proposed pillar, too, is building broad-based infrastructure for AI innovation nationwide. And this is a very comprehensive, all-inclusive pillar. This includes the acceleration of geothermal, natural gas, and nuclear permitting,

Starting point is 00:16:33 a strengthening of domestic production of critical grid components, such as through loan and loan guarantee programs, as well as supporting training and apprenticeship programs for critical energy workers, electricians, and construction workers. So really this report is all about the infrastructure required to build AI in America. But then, as I mentioned, we got from the White House its own AI action plan. Former VC Shrederam Krishna wrote, Today is a day we have been working towards for six months.

Starting point is 00:16:58 We are announcing America's AI Action Plan putting us on the road to continued AI dominance. The three core themes, accelerate AI innovation, build American AI infrastructure, lead in international AI diplomacy and security. The introduction of the report, which you can find at AI.gov slash action dash plan, is America is in a race to achieve global dominance in artificial intelligence. Winning this race will usher in a new era of human flourishing, economic competitiveness, and national security for the American people. Recognizing this, President Trump directed the creation of an AI action plan in the early days of his second term in office. Based on the three pillars

Starting point is 00:17:33 of accelerating innovation, building AI infrastructure, and leading an international diplomacy and security, this action plan is America's roadmap to win the race. Now, the actual document itself, like I said, is about 28 pages. And while it's comprehensive, it's not particularly detailed, so there will be a lot to do and figure out in the months to come. To get a sense, though, of the types of things that they're thinking about, let's go through these pillars one by one. Pillar 1 is accelerating AI innovation. The overview writes, America must have the most powerful AI systems in the world, but we must also lead the world in creative and transformative application of those systems. Ultimately, it is the uses of

Starting point is 00:18:06 technology that create economic growth, new jobs, and scientific advancements. America must invent and embrace productivity-enhancing AI uses that the world wants to emulate. Achieving this requires the federal government to create the conditions where private sector-led innovation can flourish. So what are the pieces of this? Basically, the advanced overview page has all of the bullet points, while the full report has more like a paragraph on each of them. Some of these are exactly what you'd expect, like removing red tape and onerous regulation, but a lot of them are getting into real details. One which Krishna pointed out is something that was particularly important to him, was

Starting point is 00:18:38 encourage open source and open weight AI. On that front, the report reads, open source and open weight AI models are made freely available by developers for anyone in the world to download and modify. Models distributed this way have unique value for innovation because startups can use them flexibly without being dependent on a closed model provider. They also benefit commercial and government adoption of AI because many businesses and governments have sensitive data that they cannot send to close model vendors. And they are essential for academic research, which often relies on access

Starting point is 00:19:04 to the weights and training data of a model to perform scientifically rigorous experiments. We need to ensure America has leading open models founded on American values. Open source and open weight models could become global standards in some areas of business and in academic research worldwide. For that reason, they have geostrategic value. while the decision of whether or how to release an open or closed model is fundamentally up to the developer, the federal government should create a supportive environment for open models. Then, as with each of these policy points, they have a set of recommended policy actions. In this case, those include ensuring access to large-scale computing power for startups and academics,

Starting point is 00:19:39 partnering with leading technology companies to increase the research community's access to private sector computing, and a number of others that get even more nitty-gritty. Now, this is particularly notable, because one could make an argument that this sort of runs counter to the way some of the previous policy had been oriented when it comes to China. We've continued to see more and more reductions and more and more prohibitions when it comes to things like chip exports to China, and for a while that was aligned with the point of view that you shouldn't have open source models because China can just emulate it. Now, part of what's happened, of course, is that China has raced into the lead when it comes to open models. Deepseek at the

Starting point is 00:20:14 beginning of this year was, of course, a major wake-up call. And one of the really interesting conversations is basically whether the U.S. doesn't make open models available from a cost and availability standpoint, the world will just default to Chinese models with all the attendant prohibitions that they might have around the way that history is told. Point being that the acknowledgement that open source and open weight models have geostrategic value is a big one that I think will be highly instructive to how this government thinks about AI diffusion in the years to come. Another one that I noticed, which is a theme we've been harping on here, although this is obviously a slightly different version of it, is the call to build a

Starting point is 00:20:48 an AI evaluations ecosystem. Turns out it's not just developers who are thinking more about agent evals right now. It's also the U.S. government. Pillar 2 is sort of ripped from the pages of the Anthropic Report and is called Build American AI Infrastructure. They write, AI is the first digital service in modern life that challenges America to build vastly greater energy generation than we have today. American energy capacity has stagnated since the 1970s, while China has rapidly built out their grid. One needs only look at this chart to see just how dramatic that is. They continue, America's path to AI dominance depends on changing this troubling trend. That requires streamlining permitting, strengthening and growing the electric grid, and

Starting point is 00:21:26 creating the workforce to build it all. So this is very similar to what Anthropic is calling for. Streamline permitting, developing a grid that can match the pace of AI innovation. This particular report goes farther in talking about restoring the American semiconductor manufacturing industry or reshoring it, and it also talks about training a skilled workforce for AI infrastructure. The third pillar is to lead in international AI diplomacy and security. They write, To succeed in the global AI competition, America must do more than promote AI within its own borders. The United States must also drive adoption of American AI systems, computing hardware, and standards throughout the world. America currently is the global leader on data center

Starting point is 00:22:04 construction, computing hardware performance, and models. It is imperative that the United States leverage this advantage into an enduring global alliance while preventing our adversaries from free writing on our innovation and investment. Now, it is way beyond the scope of this show, obviously, which is an AI show, to get into American foreign policy in general. But one of the great tensions for American foreign policy, particularly in this White House, is trying to navigate the line between this sort of America first withdrawal from the world and America First-style leadership of the world. This is, I believe, a central tension. And I say that without getting into any of the arguments for one side or another. What's clear,

Starting point is 00:22:42 though, is that at least when it comes to AI, this is not a U.S. government in withdrawal from its leadership role in the world. Instead, this document articulates a vision of intentional American engagement to diffuse American AI throughout the world. The first pillar here is exporting American AI to allies and partners. And to some extent, going back to that question that we had before, when we were talking about the open source and open weight models, the policy that seems to be coming together is deny China access to advanced infrastructure to train their own models at the same time as exporting American models to the rest of the world. Now, at the time of recording, this has only been out for an hour or two, and so people are just wrapping their heads around it.

Starting point is 00:23:24 Mohamed Soleiman from the Middle East Institute writes, the core message of the White House's AI Action Plan is clear, build AI infrastructure and build more of it. And to stay ahead, the U.S. must export its full AI stack, ensuring global systems are anchored on American infrastructure. Georgetown professor Ryan Fedisuk, apologies Ryan for probably butchering your name, writes, after years spent watching DC fumble tech policy, this document is different. It reads like it was written by people who understand both the technology and the stakes.

Starting point is 00:23:50 Ryan continues, the framing gets straight to the point. The United States is in a race to achieve global dominance in AI. Whoever has the largest AI ecosystem will set global AI standards and reap broad economic and military benefits. No more competing while we cooperate. Now, interestingly, on China, as much as they are a centipiece of this, Ryan argues that there is a sensible throughline,

Starting point is 00:24:10 quote, taking competition seriously without veering into hysteria, export USAI to allies before they turn to Beijing, plugging loopholes in semiconductor export controls, and countering CCP influence in international standards bodies. Ultimately, Ryan says that this feels to him grounded in reality. Like I said at the top, this report is only hours old at this point, at least for me, and so there will be a lot more to dig into soon. What's clear, though, this is a White House that sees AI as geostromes,

Starting point is 00:24:38 strategically important, something worth investing in, and something with massive implications for the shape of the future. Lots more to come, but for now, that is going to do it for today's AI Daily Brief. Appreciate you listening and watching as always, and until next time, peace.

The AI Daily Brief: Artificial Intelligence News and Analysis - America's AI Action Plan

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.