US Gov't to Get Advanced Access to OpenAI, Anthropic Models

0:00Today on the AI Daily Brief, Anthropic and OpenAI will be sharing their advanced models with the US government. Before that on the brief, OpenAI says that chat GPT usage has doubled over the course of the last year. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. To join the conversation, follow the Discord link in our show notes.

0:23Welcome back to the AI Daily Brief Headlines Edition, all the daily AI news you need in around five minutes. We kick off today with some updated numbers from OpenAI around chat GPT. The company said that ChatGPT now has more than 200 million weekly active users. That's around twice as many as it had last November. So a 100 % growth in the last nine months or so. OpenAI also reported that 92 % of Fortune 500 companies are using its products, although it's not clear to me if that means they have users with email addresses representing 92 % of the Fortune 500, which is my guess, as opposed to having contractual relationships with 92 % of the Fortune 500.

1:01And they also said that their API usage has doubled since the release of GPT-40 mini, which frankly is an even more telling statistic, given that represents a more sophisticated usage that's probably more integrated deeper into the enterprise. In a statement for Axios, CEO Sam Altman said, people are using our tools now as part of their daily lives, making a real difference in areas like healthcare and education, whether it's helping with routine tasks, solving hard problems, or unlocking creativity. As with so many things in this space, the way that someone interprets this news is pretty much a Rorschach test.

1:32Is it a sign of the extreme growth of the space, that you now have this product that didn't exist even two years ago, that 200 million people are using every single week? Or does the fact that it got to 100 million people who tested it in the first five weeks after release, back in November 2022, and then took a year to double its usage to 200 million people, suggest that the rate of growth is slowing down? Is it possible that both are true? That there was an initial novelty explosion, a period of equilibrium, And now we've moved into more of a period of steady growth. And as we see from that API usage, perhaps even more sustainable growth.

2:05You will probably find it unsurprising, given where I sit professionally, that I think that a product that didn't exist two years ago, getting to 200 million people using it every single week, is telling of just how significant this category of technology is. Speaking of OpenAI, it does seem like that$100 billion valuation investment round that we've been talking about is the same tender offer-focused round that they've been working on for some time now. Now, maybe the scope has expanded a little bit so that it's not just a tender offer of employee shares, but the point being that this isn't some new round that we hadn't heard anything from.

2:38But then again, it might become that. The Wall Street Journal is reporting that Apple, who is of course integrating ChatGPT into iOS, is in talks to participate in this round. Bloomberg also reported that Nvidia has discussed joining the funding round as well. Now, this isn't that surprising. Before being spooked by potential antitrust concerns, Apple's Phil Schiller was scheduled to join OpenAI's board. What's more, with ChatGPT representing such a significant place in the generative AI sphere, you would kind of expect the giants to try to at least get a little piece of that action. Now, on the growth side, OpenAI wasn't the only company reporting some significant growth.

3:14Mark Zuckerberg took to Threads to announce, quote, Llama is growing even faster than I expected. Almost 350 million downloads, over 20 million in the last month, and a 10x jump in monthly usage since the start of the year. excited to share the next set of updates and Lama models soon. He also said Meta AI now has more than 400 million monthly actives and 185 million weekly actives across our products. Now, the first statistic I think reflects growth in developer usage of Lama after the release of the Lama 3 models, and probably even more after the release of the 3.1 models, including 3.1.405b.

3:47The 400 million monthly active number I think has to do more with the integrated AI features into experiences like Instagram and WhatsApp. Still, to the extent that those types of experiences are people's gateways into AI, there certainly seems to be a lot of people touching them. Now, some people are saying that this is a jump-the-shark moment. Variety reports that Oprah Winfrey is set to host an AI and the Future of Us ABC special featuring luminaries like Bill Gates and Sam Altman. It is set to air September 12th and promises to, quote, provide a serious, entertaining, and meaningful base for every viewer to understand AI and empower everyone to be part of one of the most important global conversations of the 21st century.

4:53The promo continues. side of the conversation, at least for a mainstream audience. But it is notable that they're inviting Tristan Harris and Aza Raskin, who are the founders of the Center for Humane Technology, who are much more on the AI safety side as part of the discussion. Come September 13th, we will have more on this. A project that has a lot of people excited this week is MAGIC, which has announced a model with ultra-long context, specifically 100 million tokens. Now, the team had announced this work previously, but added some additional updates, including a partnership with Google Cloud to create two supercomputers and a new$320 million investment round from investors including Eric Schmidt, Jane Street, Sequoia, Atlassian, and more.

5:32One interesting note, which might give you a sense of where the commercialization of AI is right now, they write, while the commercial applications of these ultra-long context models are plenty, at Magic we are focused on the domain of software development. If you've been in and around AI Twitter recently, you'll know that everyone is just talking about Cursor. Although the AI coding domain has always been an anticipated early adoption sphere, it seems like it is really taking hold as the first industry where people believe that it will be completely reimagined in the context of AI support. Along those lines, another AI coding platform, Codium, has announced a$150 million Series C.

6:05The company has become the latest AI unicorn with a$1.25 billion valuation. In under two years, the company has grown from zero to 700 ,000 active users and a paid enterprise business from zero to eight figures in ARR. Lastly today, Yale has announced$150 million to support leadership in AI. They will commit more than$150 million over the next five years to, quote, support faculty, students, and staff as they engage with artificial intelligence. Said Provost Scott Strabell, Yale has long been at the forefront of AI development and research, and our leadership continues to be necessary as this technology evolves and endures.

6:39To fulfill the university's mission to improve the world and prepare the next generation of society's great leaders and thinkers, we must explore, advance, and harness AI for its benefits while providing ethical, legal, and social frameworks to address the challenges it poses. So within that, some of the commitments that they're making, including expanded research infrastructure, developing a custom clarity platform, offering a walled-off version of ChatGPT 4.0 for Yale faculty, students, and staff, adding faculty positions around these areas, and seed grants for new curriculum. Overall, just another sign of how people are building this out for the long term.

7:09For those of you who are still on the AI is Just Hype train, I would ask you whether Yale ever created a$150 million center for NFTs and the metaverse. For now, though, that is going to do it for today's AI Daily Brief Headlines edition. Next up, the main episode. Today's episode is brought to you by Fractional. When we wanted to build an AI-powered feature of Superintelligent, our AI tool finder, I went straight to Fractional. The Fractional team is a group of senior engineers in San Francisco working on some of the most exciting projects in applied AI. working with them is basically like hiring an absolute top flight AI engineering team, but in a way that you can customize exactly for your particular needs.

7:49Like I said, that AI tool finder feature that we built with them is already a key part of the super intelligent platform. And we're working on something new as well. Fractional works with everyone from startups to the fortune 500. And to request a free consultation, you can go to fractional.ai. If you want help identifying and building AI projects for your business, then I highly recommend you hit pause, open a web browser, and go to fractional.ai to request a free consultation.

8:36Venice is an alternative. Venice is a powerful AI app for text, image, and code generation that respects you as a sovereign individual and believes privacy and free speech are not only human rights, but necessary for civilizational advancement. Private, permissionless, and uncensored, you can try it for free without an account. AI Daily Brief listeners receive a 20 % discount on Venice Pro. Visit venice.ai slash NLW and enter the discount code NLWDailyBrief. That's NLWDailyBrief. All one word. Today's episode is brought to you by Super Intelligent, which is, of course, our platform that helps you learn how to use AI tools and perhaps even more importantly, gives you ideas on the best use cases that are actually going to help you achieve whatever it is you want to achieve.

9:21To recognize the end of summer and back to school slash back to work, we are running our best promotion ever. When you sign up for Super Intelligent between now and the end of August, using code SOBAC, your first month will be 100 % free. The platform features over 600 fun, highly practical AI tutorials that get you using AI fast and with an eye to actually transforming how you get things done. We've just launched Super for Teams, so if you have a group of people at your company that want to figure out how to use AI together, I highly suggest you check it out. But for those of you who are using Superintelligent as an individual, once again, if you sign up for super intelligent between now and the end of the month using code SOBAC.

10:00You will get your first month 100 % free. Go to besuper.ai and check it out today. Welcome back to the AI Daily Brief. The discussion for the last couple of weeks has just been nonstop regulation. Of course, the big focus has been California's SB 1047. The controversial bill was passed by the State Assembly earlier this week, then ratified yesterday by the Senate, and is now moving on to the desk of Governor Gavin Newsom for signature or veto. One of the companies against the bill has been OpenAI. The company had said, for example, that they were worried that it would slow down progress, and they also argued that the rule should be federal, not state by state.

10:38Said Chief Strategy Officer Jason Kwan, A federally driven set of AI policies rather than a patchwork of state laws will foster innovation and position the U.S. to lead the development of global standards. As a result, we join other AI labs, developers, experts, and members of California's congressional delegation in respectfully opposing SB 1047. Interestingly then, people were very interested when on Thursday afternoon, OpenAI CEO Sam Altman tweeted that the company had reached an agreement with the U.S. government around pre-release testing of their most advanced models. He also got a subtweet in about SB 1047.

11:10Altman tweeted, We are happy to have reached an agreement with the USAI Safety Institute for pre-release testing of our future models. For many reasons, we think it's important that this happens at the national level. U.S. needs to continue to lead. This is not the first time that Altman has made the argument of the importance of U.S. leadership. In July, he published an op-ed in the Washington Post called Who Will Control the Future of AI? A democratic vision for artificial intelligence must prevail over an authoritarian one. Altman basically argued in that piece that it's either going to be the U.S.

11:39and the democratic nations of the world, or it's going to be China. You might or might not remember that Altman had also talked about the fact that in a different world, OpenAI would have actually come from the government. Back in 2023, another Post reporter was hanging out at the OpenAI headquarters in San Francisco when Altman said, OpenAI should have been a government project, right? He continued, In a different time, it would have been. At a minimum, we should have gotten government funding, which we couldn't. We tried and failed. That piece continues. Altman said that in 2017, he pitched officials at the White House, the Defense Department, and the

12:21The point of this is that Altman clearly has a sense of the national security and geopolitical significance of this change. But back to this agreement, the agreement is technically between Anthropic and OpenAI and the USAI Safety Institute. The USAI Safety Institute is housed within the National Institute of Standards and Technology and was one of the things that come out of Biden's executive order on AI. There are about page reads, Our efforts will initially focus on the priorities assigned to NIST under President Biden's executive order on AI. The Safety Institute will pursue a range of projects each dedicated to a specific challenge that is key to our mission.

12:55These will initially include advancing research and measurement science for AI safety, conducting safety evaluations of models and systems, and developing guidelines for evaluations and risk mitigations. Now, as to this deal, they call this a first-of-a-kind agreement between the U.S. government and industry that will, quote, enable formal collaboration on AI safety research, testing, and evaluation. Details are scant, but the core is simple. The U.S. AI Safety Institute is going to get access to major new models before they are released publicly. The question, of course, is whether they will have the ability to actually delay the release of those models should there be a disagreement around safety implications.

13:28How much influence will the AI Safety Institute and, by extension, the U.S. government have on what type of guardrails are put in place. For some, this was a positive advancement. Zvi Maushiewicz, who is one of the most thoughtful writers when it comes to AI safety issues, wrote, highly welcome news that they and Anthropic will now do this. But he also added, amazing to see the flood of absolutely vile responses. True Ventures partner Gus Koldabella called out the state regulation issue. Gus writes, whether you like or dislike what Sam announced, he's subtly making an important point about state regulation of AI, putting aside SB 1047's destructiveness towards open-source AI and AI innovation in general, which has been overwhelmingly demonstrated by many others, its passage could start a state-by-state legislative arms race, and that's good for nobody but the lawyers.

14:08Imagine if 10, 30, or even 50 different states enact their own AI regulatory regimes, each with different expensive requirements, periodic reports, liability standards, and new bureaucracies. Resources that developers could spend on the kind of innovation we want are instead soaked up by compliance, if companies can stomach the liability risk to begin with. It's a strategic imperative to have the U.S. lead in AI. We should not stumble into a patchwork of state laws that discourages innovation before it begins. A16Z partner Martin Casado, who's been one of the loudest opponents of SB 1047, responded and wrote, Yup, also it's much more aligned with how we've treated state-of-the-art systems in the past.

14:41Gus used that as a chance to clarify, My post is certainly not advocating a rush to regulate it at the national level. God, no. Just a call to avoid the privacy law of vacation, i.e. different and often irreconcilable state-by-state standards of AI. However, that's basically where the positive takes ended. George Hotz wrote, Tell me how it feels to have your company infiltrated by the deep state. What do they threaten you with? AI entrepreneur Bindu Reddy writes, The government testing AI models. Yeah, sure, they are the experts. Expect delays of two to ten years. Maria Simplified writes, POV, it's 2040 and two government agents are knocking on your door.

15:14You've used algebra in an unsafe way, and now you must surrender all your mathematical formulas before someone else gets exposed. There are about a million others that have that same tone. Then, of course, there were the joke posts, like this one from Matthew Berman, who shared a picture of a sad Steve Carell, saying me when I know the feds have seen GPT-5 and I haven't even tried GPT-40 voice. Boy, do I relate to that one as I sit here with this AI podcast and this AI company and no advanced voice mode. Others pointed out that this seems to add credence to the big piece that Leopold Aschenbrenner, formerly of OpenAI, wrote earlier this year.

15:45One section that was highlighted from Aschenbrenner's piece, somewhere around 26, 27 or so, the mood in Washington will become somber. People will start to viscerally feel what is happening. They will be scared. From the halls of the Pentagon to the back-run congressional briefings will ring the obvious question, the question on everybody's minds. Do we need an AGI Manhattan project? Slowly at first and all at once, it will become clear this is happening. Things are going to get wild. This is the most important challenge for the national security of the United States since the invention of the atomic bomb.

16:11In one form or another, the national security state will get heavily involved. This inevitably is going to cause the U.S. to fall behind China. Bindu Reddy again writes, all this premature AI safety stuff will cause U.S. to give up its lead. China is already on par with us and may soon get ahead. Quen, Kling, and the Yi models are very good, and if the US slows down, the world will be using their models. We may have another TikTok-like situation where they, not us, are number one. Notably on the same day, Quen announced that its latest model was better than GPT-40 based on the metrics, adding some amount of credence to that catch-up sort of sense.

16:43Then again, just to add a little extra wrinkle to this, The Economist recently wrote a piece asking whether Xi Jinping is an AI doomer. The interesting quote, China has its own AI doomers and they are increasingly influential. Yijang of the Chinese Academy of Sciences believes that AGI models will eventually see humans as humans see ants. In other words, friends, it is just going to get weirder from here, so stay tuned. Thanks as always for listening or watching, and until next time, peace.

17:22Thank you.