Why Everyone is Angry About the AI Safety Board

0:00Today on the AI Breakdown, we're talking about everyone's angry response to a new Department of Homeland Security AI Advisory Board. Before that on the brief, Apple in talks with OpenAI about the iPhone. 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 Discord, and our newsletter. One quick note before we get into the show today, if you have been considering diving into the wild, crazy AI learning world of super intelligent, and you want to try it in May, I am running a special cohort that I will be personally managing.

0:37Basically every week I'll send a bunch of recommendations for tutorials you should be watching and projects you should be doing, and we'll also have a private dedicated Discord channel where I'm happy to hear about not only what you're trying to learn in AI, but your own projects as well. You can go to bsuper.ai and when you sign up, use code NLWMay. You'll get$5 off your first month and be automatically included in that cohort. Registration closes for that at the end of this month, so basically tomorrow on Tuesday. I would love to see you there. For now though, let's get into some Apple rumors.

1:06Welcome back to the AI Breakdown Brief, all the AI headline news you need in around five minutes. If you're not getting it yet, let me explain what the next few months before Apple's worldwide developer conference are going to be like. Every week, we're going to get some new rumor about something that is coming or some partner that Apple is negotiating with, some new model that they've released on research, some new feature that someone reports is coming. The curiosity around how Apple is going to compete in the generative AI space is, in other words, reaching an absolute fever pitch. And the current discussion in this area is around OpenAI powering certain iPhone generative AI features.

1:43Now, you may remember, back about a month ago, Bloomberg's Apple whisperer Mark Gurman reported that Apple was in talks with Google Gemini about powering iPhone AI features. Since then, we've talked extensively about what that might look like, why Apple might cede some of that territory to Google, with much of the working conventional wisdom, pointing to the idea that as Apple was trying to control the AI in certain experiences, such as using their own models for, for example, a new and more powerful Siri, there could be certain types of experiences where they just needed to have someone else powering them, at least in the short term.

2:16At the time that Gurman reported this Google conversation, he made it seem like those conversations were a lot farther along and more active than another conversation that he had heard about, which was between Apple and OpenAI. Apple and Google, of course have an extensive financial relationship currently. Specifically, Google has been paying Apple billions of dollars a year to make its search engine the default option on Apple devices like the iPhone. At the time that Gurman reported these talks, it wasn't clear if Apple would be paying Google, if Google would be paying Apple, or any of the other sort of branding issues that might come up with a deal like that.

2:48Well, at the end of last week, Gurman once again came back with a new piece titled, Apple intensifies talks with OpenAI for iPhone generative AI features. As always, this is according to people familiar with the matter, and Gurman writes that Apple has renewed discussions with OpenAI about using their technology to power some features coming to the iPhone later this year. And apparently this was a reopening of a dialogue. Quote, Apple had talked to OpenAI about a deal earlier this year, though work between the two parties had been minimal since then. What's more, he also reports that this does not mean that a Google deal is off the table, and that according to these sources, not only has Apple not made a final decision on which partners it will use, There's no guarantee that it will use any of them.

3:27And if they did decide to, there's no guarantee they would just use one partner. In other words, there is still a lot to be figured out. However, what is clear and what was newsworthy is that these open AI discussions had picked back up in a significant way back from that report a month ago. Daniel Newman summed up a lot of the conversation I saw on Twitter around this, writing, Very interesting if Apple goes with open AI. One, real implications on the Google relationship. Two, what does it say about its own AI efforts in building LLM and smaller edge models? Three, will Apple really want to be more tightly tied in with Microsoft via OpenAI?

3:58Four, Apple has to do something and increasingly it doesn't look like it will be in-house. Now, I think it's a little too early to fully determine any of those questions. Certainly, it does seem like the Microsoft thing might be a barrier. But my strong instinct is that these conversations with OpenAI and Google do not represent Apple abandoning in-house efforts. It seems like them being realistic about what they can do in the time they have allotted, and knowing that once they crack the seal on AI features for the iPhone, they have to achieve parity with some of the other things out there being offered.

4:27I would be very surprised if the long-term trajectory doesn't continue to have Apple seeing itself owning the end-to-end experience. Now, of course, part of the reason that Apple might have to do one of these partnerships is that getting in the AI game is very, very expensive and very, very complicated. Yesterday, in response to a tweet from the Whole Mars Catalog, which said Tesla spent$1 billion on training compute in Q1 and doubled their compute capacity in the process, Elon Musk responded, Tesla will spend around$10 billion this year in combined training and inference AI, the latter being primarily in car.

4:57Any company not spending at this level and doing so efficiently cannot compete. This, of course, may remind you from one of our episodes last week about Grok and their reported$6 billion fundraise. Now, interestingly, there are a lot of articles this week about exactly this issue. The New York Times, for example, wrote a piece called AI Startups Face a Rough Financial Reality Check. The table stakes for small companies to compete with the likes of Microsoft and Google are in the billions of dollars. and even that may not be enough. They point to Inflection, whose team has almost entirely gone to Microsoft, Stability, who has had, of course, lots and lots of leadership turnover, and Anthropic basically being in a never-ending rush to raise more money.

5:34Another related piece in the Times from this weekend, in Race to Build AI, tech plans a big plumbing upgrade. The spending that the industry's giants expect artificial intelligence to require is starting to come into focus, and it is jarringly large. Of course, that came after Meta wiped out a couple hundred billion dollars off its market cap by telling investors that it was going to increase the amount that it was spending on AI this year by about$5 billion. The question, I think, which again, I explored on that episode about Grok, is how much the open AIs and anthropics of the world can compete if companies like Meta keep being willing to give away models that are close to the state of the art.

6:08Will that change the calculus for OpenAI's fledgling enterprise business, which is at a $2 billion or so revenue run rate? Well, whatever the case when it comes to their business model, OpenAI continues to pump out new features. ChatGPT's memory function is now widely available. You can do things like create a specific indicator for a type of prompt and ask ChatGPT to put it in its memory notes so that anytime it sees that indicator, it can execute that specific prompt. This is one of those features that will be insane to think that we didn't live without it now that it's here. OpenAI today also announced a licensing deal with the Financial Times.

6:41There continues to be a little bit of narrative confusion, even when announcing these deals, around how much it's about officially licensing material to the training of ChatGPT versus new features that interact with the Financial Times and other outlets in some way. Sam Altman, for his part, has said that deals like this are much more about the ways in which ChatGPT can interact with the existing news system rather than getting access to some data that they need for their model to work. In this case, for example, FT writes, The agreement allows ChatGPT to respond to questions with short summaries from FT articles with links back to FT.com.

7:12This is the fifth deal of its type that OpenAI has closed this year, with the others being the Associated Press, Axel Springer, Le Monde, and Spain's Prism Media. That, however, is going to do it for today's AI Breakdown Brief. Next up, the main AI breakdown. Attention, AI Breakdown listeners. Consensus 2024 marks the 10th gathering for all things crypto, blockchain, and Web3. However, importantly, this year's agenda will also dive deep into AI-driven transformation. And the speaker lineup includes the leading minds and innovators at the forefront of this digital renaissance. Don't miss the Consensus AI Summit to cut through the hype to find where true transformation and opportunity lie.

7:50Listeners to this show can get 15 % off registration with the code AIBreakdown. Visit consensus2024.coindesk.com to learn more. Some of the folks who will be at Consensus this year include Guillaume Verdun, aka Beth Jezos, founder and CEO of Xtropic, as well as spiritual leader of the accelerationist movement, Neil Stephenson, co-founder of Lamina One, and Brendan Icke, the CEO of Brave Software. Again, go to consensus2024.coindesk.com to learn more and get 15 % off registration with the code AIBreakdown. Today's episode is brought to you by Notion. Notion is one of a very small handful of tools that is at the very center of every new project I begin.

8:29That includes the AI Breakdown, Superintelligent, previous things that are long gone in the past, Notion is always a big piece of it, and that's because Notion combines your notes, docs, and projects into one space that's simple and beautifully designed. They have now put AI directly at the center of that platform in ways that are integrated to make it actually really useful. You can use it to help you brainstorm, use it to help you draft things, or use it for intelligent search across all of the documents your team has produced, which can be an incredibly time-saving feature. Notion is used by over half of Fortune 500 companies, and teams that use Notion send less email, cancel more meetings, save time searching for their work, and reduce spending on tools.

9:08Right now, for AI Breakdown listeners, you can try Notion for free when you go to notion.com slash AI Breakdown. That's all lowercase letters, notion.com slash AI Breakdown. When you use our link, you're supporting the show. Once again, that's notion.com slash AI Breakdown. Welcome back to the AI Breakdown. The artificial intelligence space is not without controversy, as all of you listening will well know. Within the AI space, there are extreme ideological divergences between many of the people who spend all their time in this area. On one far end of the spectrum, we have the safety and pause folks.

9:44On the other end of the spectrum, we have the accelerationists. And then, of course, there are the folks who want to jump up and down on that spectrum because they think that AI is just stealing art, knowledge, creativity, etc. It was a true rarity then, when at the end of last week, the Department of Homeland Security announced the establishment of its first-ever Artificial Intelligence Safety and Security Board that every one of these groups, and representatives of individuals from all of these groups, universally condemned the 22 inaugural members who were added to the committee. Truly, we are coming together in hatred today.

10:15And obviously, we're going to talk a little bit about what this says about the state of the discourse around AI safety, as well as the state of the discourse around regulatory capture in Washington, D.C. First of all, who is included? On the one hand, there are those who are leaders of startup AI labs, Sam Altman, the CEO of OpenAI, Dario Amadei, the CEO of Anthropic, more established CEOs like those from NVIDIA, IBM, Microsoft, and Adobe. We also have some academics, like Nicole Turner-Lee, a senior fellow and director of the Center of Technology Innovation at the Brookings Institution. Alexander Reeve Givens, the president and CEO of the Center for Democracy and Technology.

10:49A few civil rights advocates, such as Damon Hewitt, the president and executive director of the Lawyers Committee for Civil Rights Under Law. And Maya Wiley, the president and CEO of the Leadership Conference on Civil and Human Rights. And then we have some more random additions. We have Kathy Warden, the CEO and president of Northrop Grumman, who at least is in this whole military-industrial-government contractor area, so maybe has expertise in that way. But then you also have the CEO of Delta Airlines, the mayor of Seattle, the governor of Maryland, and a handful of others. So first of all, let's talk about the purpose of this group, and then we'll get into some of the critiques.

11:23The Wall Street Journal writes, The AI Safety and Security Board is part of a government push to protect the economy, public health, and vital industries from being harmed by AI-powered threats. Working with the Department of Homeland Security, it will develop recommendations for power grid operators, transportation service providers, and manufacturing plants, among others, on how to use AI while bulletproofing their systems against potential disruption that could be caused by advances in the technology. Now, one of the big things that you'll see a lot of critiques focus in on is represented here by Saliferous Games, who writes, This is such blatant conflict of interest.

11:55They're letting the AI industry write the laws for the AI industry. Ed Newton-Rex, who left Stability AI to start new approaches to more rights-ready AI training writes, why does the new AI Safety and Security Board have so many members with clear vested interests and rapid unregulated AI adoption? And so few who have AI expertise without those vested interests. Now one thing that should be noted here is that this is not a generalist government group to recommend on generalist AI policy. This is specifically focused on infrastructure challenges that come up with AI and the potential for infrastructure attacks.

12:27The WSJ interviewed Alejandro Mayorkas, the Secretary of Homeland Security, who said that embedding AI technology within the nation's water facilities, transportation systems, and banks presented tremendous opportunity to improve service to benefit Americans' daily lives, but also carried significant risks. He continued, A failure to deploy AI in a safe and secure and responsible manner when it comes to critical infrastructure can be devastating. Now, I don't think that this entirely undermines that critique of the vested interests, but I do think it's worth noting what the mandate of this group is and what it isn't.

12:57Now, of course, with the names presented here, they're very clearly positioning this as a very important agency, otherwise you wouldn't get this sort of participation that they're getting. It seems like it was pretty clear to the Department of Homeland Security that this was going to come up as a critique. Again, from the journal, Mayorkas said he wasn't concerned that the board's membership included many technology executives working to advance and promote the use of AI. Said Mayorkas, they understand the mission of this board. This is not a mission that is about business development. Now, when it comes to what the board will actually do, they have their first meeting coming up this month in May, and they are apparently expected to convene quarterly.

13:31The idea of vested interests actually had a bunch of different flavors in the critique. Cole Roberts writes, looking like the regulatory capture monstars. Dr. Gingerballs writes, regulatory capture working exactly as intended. What a sham. Janelle Jumalan writes, let's have oil companies be on the board for green energy is what this is. Okay, so we have critiques that are a generalist vested interest concern from people who don't like AI that much, and then a regulatory capture critique from people who are in AI, which of course is closely related to the idea articulated by Accelerate Harder, who wrote, man, I'm so irritated by this list.

14:05Is there even one person here to take the side of open source AI? What a joke. A16Z's Martin Casado writes, almost no representation from startups or open source. Of course, this is related to that regulatory capture idea and is part of the specific concern about it. Yet another category of critique is what people are interpreting as purposeful snubs. Mihai Doho writes, DHS has deliberately not included the AI open source pioneers like Musk and Zuck. They don't want the people to own AI, they want to regulate AI so it's centralized power. We can't allow AI to belong to an elite. It needs to be free and open to all of us like the internet.

14:39Now, it was very notable last year when the White House started inviting CEOs to talk about AI that they did not include Mark Zuckerberg. When asked about the snub, the press secretary had a very snarky statement basically saying that this meeting was for people who are influential in AI only. Now, given how much Meta and Zuckerberg have shaped the open source side of this development, it's hard to read it as anything other than they do not like Mark Zuckerberg. Now, ultimately, this is just one random committee. How much they're actually likely to do, I think, is a very reasonable question. And so in that way, I think what's most interesting about it is these reactions and what they reflect about the state of the conversation.

15:16It is clear that from multiple sides, there is a significant concern with a very small handful of leading big tech industrialists being the ones to shape AI policy, even if those concerns are for entirely different reasons. Still, in the short term, I have to agree with Andrew Curran, who writes, the amazing thing about that AI board post was that for a few hours, everyone on my feed forgot they were supposed to be fighting. Safety, ACC, anti-AI artists, open source, all in near unison turned to look at this new thing that had appeared, and in one voice, condemned it. I will certainly be watching to see if anything more comes of this, but for now, that is going to do it for the AI Breakdown.

15:53Until next time, peace.

16:06Thank you.