The AI Daily Brief: Artificial Intelligence News and Analysis - The 9 Most Important AI Stories from October

Episode Date: November 1, 2024

October saw major developments across the AI industry, from the White House prioritizing AI for national security to new funding for OpenAI. Key updates included Google’s shift towards nuclear energy, Anthropic’s new AI computer-use capabilities, and a national conversation on California’s SB 1047. Join us for a breakdown of the month’s most impactful AI stories! Concerned about being spied on? Tired of censored responses? AI Daily Brief listeners receive a 20% discount on Venice Pro. Visit ⁠⁠⁠https://venice.ai/nlw⁠⁠⁠ and enter the discount code NLWDAILYBRIEF. The AI Daily Brief helps you understand the most important news and discussions in AI. Subscribe to the podcast version of The AI Daily Brief wherever you listen: https://pod.link/1680633614 Subscribe to the newsletter: https://aidailybrief.beehiiv.com/ Join our Discord: https://bit.ly/aibreakdown

Today on the AI Daily Brief, a conversation about the most important AI news from October. The AI Daily Brief is a daily podcast and video about the most important news in AI. To join the conversation, follow the Discord link in our show notes. Hello, friends. Today is, of course, the last day of October. Happy Halloween. And happy Mariah Carey season to those who celebrate. Now, for our purposes today, we are doing something a little bit different. October was a pretty consequential month. We had really interesting news on the AI and political fronts.

tons of funding news, some model developments. And so joining me to count down and go through all of the most important stories is Tiana Baker-Taylor. Tiana is the co-founder and COO of Venice AI. And I hope that you will join us as we discuss what shaped AI this month and what we see coming up next. All right, Tiana, welcome back to the AI Daily Brief. How are you doing? I'm great. NLW. How are you? Things are good. It's Halloween season. Over in Cryptoland, it's October, which is finally kicking in. Actually, October in Cryptoland is also intersecting with AI, with mining stocks, where Bitcoin miners who have pivoted to fast infrastructure buildouts for AI are kind of doing really well.

So, you know, it's all happening. It's a great time. It's my favorite time of year. And so I thought it would be fun to have you back on. And basically, you know, we're coming up to the end of October. It was a fairly active, significant month. And so I thought what would be fun is to sort of recap some of the biggest. stories in roughly chronological, although not exclusively chronological order. So we have about

eight or nine stories that I think kind of tell, you know, broadly speaking, what was happening in the markets this month. Awesome. Let's do it. Okay, cool. So I want to kick off with, we'll actually kick off with something sort of like a bookended government-related intersections with crypto. The big story and the big discussion at the end of September was around the SB 10, right, this California regulation, which was passed by the California legislature and at the very end of September vetoed by Governor Gadsden Newsom. We had talked a little bit about that. I think the last time you were here, which was sort of on the early side. It was pre-voting. You know, did, let's just start with, did, was your expectation that Newsom was going to veto or were you not sure? And what was your

take on sort of that as a, as a moment, both in terms of the bill itself, but also in terms of terms of sort of the broader conversation around AI safety. Yeah. So, you know, I've spent a lot of time with policymakers over the years, and you can generally kind of get a sense, especially about something as discussed as this, which way they're leaning. And I honestly had no idea. I think this could have done either way. You know, a veto is obviously, you know, a governor's prerogative, but, you know, it's a big deal, right, to basically tell your state legislature, no, you're wrong. I'm not going to sign this in. And it's also a bit of a moratorium, right? There's nothing to say that another bill can't come back, potentially under another governor. So, you know, it's a political move for sure. I think what

it signals, though, is that, you know, from a California PLC perspective, to have such stringent regulation placed on a burgeoning industry, which brings a lot of GDP to California is problematic, right? And they, you know, there was a lot of lobbying around this bill. There was a lot of problems with this bill. And I think that when you lump something together with safety, you start to almost kind of gaslight people into, like if you're not for, you know, protecting people, which this bill is going to do, then, you know, you're being kind of anti-American. And when you package up some good initiatives with some really, really crappy or not well considered or thought through policy,

like you just don't have a choice, right, but to kind of quash a bad bill. So I got to say, I was a little surprised that he did, but I'm very glad that he did. Yeah, you know, it's interesting. So there's a couple things that I thought were. It's also an election year, right? So, you know, Democrats don't need to be unpopular.

Yeah. With industry right now. It's interesting. So I think that there were a couple things that I thought were really interesting about this whole process. One was that I actually think that while the dialogue in certain circles and at certain times got kind of rancorous as is to be expected, broadly speaking, I kind of think that everyone came out of this discussion with a better understanding. like to the extent that a bill's job is to create a more engaged and informed, you know, electorate and populace, I think that this absolutely happened here. One of the things that was really notable is that, again, hold aside maybe the sort of the extreme

voices on either side, there were lots of pieces of the bill which were totally non-controversial and which like actually clearly create a path for consensus, you know, consensus areas. where you can build foundations upon actually getting to compromise an agreement, right? Whistleblower protections was a really obvious one, which is like, it's not an insignificant thing to say that that's a really important. That is a very strong safety mechanism or safety valve, you know, for some of these issues. So there's a lot of sort of positive pieces of that. I think that a couple of things that stood out in terms of where considered opposition really came from. That wasn't just sort of like the political calculus of like industry in California,

which is obviously Newsom has to think about. One, I think that people had a really hard time with California dictating the shape of legislation for the rest of the country. And I think that one of the clear sort of follow-ups of this is, you know, this has to be a national conversation. It can't just be a state-by-state thing. A second thing that came out pretty clearly is that even among Democrats, certainly among Democrats outside of California, the idea of, of, you know, the idea of, you know,

spending as much emphasis regulating theoretical risks as risks that are clear and known is very, it was a very hard pill to swallow. It felt in some ways like out of sequence for how people might imagine. You know, my perception of people's response to this wasn't even that they were unwilling to consider those risks or to just totally write them off as sci-fi and for the future. It felt it's almost like because it was so imbalanced towards those risks as opposed to current things, it almost sort of, it felt like it kind of like revealed its hand to some people as that being the main objective, you know? So I thought it was an interesting, an interesting process overall. I think if I were in the AI safety space, I would not be, I mean, obviously,

you know, you're going to be frustrated if your bill loses, you know, in this way. But I, I don't think that it's sort of some catastrophic defeat for, you know, the principles that that cohort is thinking about. I just think it's, it wasn't going to be this bill. It wasn't going to be. the way that it came through. And it's got to be a national conversation. We've seen this happen often in the crypto space, right? Where you have what would otherwise be considered a pretty decent bill that is packaged up with some stuff that is either too extreme and one end of the spectrum is a bit of a false flag, right? Doesn't necessarily speak to the rest of it. Or you're trying to do too many things at once, right? And I think some of the bigger crypto bills that have

gone through, that's where they've fallen over, right? Like Cynthia Lopez's bill is a great bill, but there's a lot in it, right? And I think when you have a new technology like this, breaking it down into little pieces, right? How are you going to regulate stable coins? How are you going to regulate trading platforms, right? Is potentially an easier way to go about implementing different policy for different market actors or activities, right? This bill was doing a lot. And I would have, if it hadn't to gain kind of so much momentum, put it into the bucket of, you know, a messaging bill.

And when it first came out, that's how it felt to me, right? This is designed to start a conversation, not designed to become legislation. So as it gained momentum, I mean, it was really an interesting bill to watch. But ultimately, I think pieces of that, legislation should probably be reintroduced without some of the again kind of false flag stuff that you know was sitting around the edges that potentially yes we're regulating for risks that have not yet been realized and I also think that maybe the AI bill in the European Union is a bit of you know a canary in the coal mine to other legislators in other places of the world

where you have companies in Europe that are basically either leaving or not releasing their new innovation to the countries that they live in, you know, Mistral being one of them. So, you know, I think a lot is happening really fast. And this bill was really unexpected. The whole trajectory and journey of it. I think that you're right that one of those sort of when industry complains, you know, we won't be there. We won't service that market if this goes through. I think in general, historically speaking, there's a, you know, I'm going to call BS kind of attitude towards that.

It's like, you're going to be wherever there's a market. So the fact that that's actually how it's playing out in Europe, I think, is more, more notable than perhaps it might otherwise have been. Yep. So, okay, so let's contrast that. Let's pair that with something that was actually just from the end of last week, which was the White House released a national security memo, basically putting AI in the context of national security. What was your perception of this? You know, was this a thing that we expected to see? Was it sort of, you know, out of nowhere?

And, you know, what did you think as you were kind of digging in and looking at, you know, the substance of this thing? Yeah. So it was something we expected to see. So the memo was a requirement of the executive order that was issued last October. So we knew that something. was going to be produced. I think what's notable is the depth and breadth of the amount of issues that this memo covered.

It was over 40 pages. And it was really comprehensive in its articulation of the administrations and the United States is, at least at this moment in time, national security strategy and public policy toward AI in general. And the executive order really focused on kind of, you know, AI governance and risk management, but this really dug into the next generation of AI systems, you know, that aren't necessarily, you know,

things like stuff that the military had been using, AI enabled applications like face and voice recognition, that kind of stuff. And it really focused on frontier models again. And so I think what's compelling is that the explosion of importance of these frontier models, I think, is really led by the popularity of Chat-Tbtbt. And I think we're going to talk about some of their numbers later. But it's really created a shift in what is considered kind of pressing and urgent from a national security perspective. And it gives some definitions, which I think is kind of helpful. So the memo defines frontier models as general purpose AI systems near the cutting edge of performance as measured by widely accepted publicly available benchmarks or similar assessments of reasoning, science, and overall capabilities.

So that's super broad. But again, it's kind of bringing into scope all of the technology that is in the hands. of people that we're playing with today. And I think the other statement that I took a way that I thought was interesting around why this is so pressing for this administration was, let's see, what did they say? Recent innovations have spurred not only an increase in AI use throughout society, but also a paradigm shift within the AI field. And this trend is most evident with the rise of large language models,

but it extends to a broader class of increasingly general purpose and computationally intensive systems. And so the U.S. government is urgently considering how this current paradigm shift could transform national security. The other thing I thought was interesting was when you're reading through these 40 pages, like, who are they talking to, right? Who's the intended audience for this? And certainly, you know, federal agencies that oversee or have scope are. around national security and obviously AI. So it gives them some specific actions that they might need to be taking.

It is also talking to AI companies that are based in the US around maintaining AI leadership and what the government is willing to do for them and what the government wants from them. But there is also quite a lot of focus on what the US expects from its allies and what the US is watching. for from its adversaries and competitors, namely China. Yeah. So one of the things that really stood out to me, you know, and again, this is the focus of it,

but there was such a clear, you know, this is, you can boil this down or at least parts of this down to staying ahead on AI is officially a national security priority, right? This is a, you know, competitiveness and sort of owning AI innovation is. is a top priority. And it's interesting because this is sort of, it echoes language that we saw in Chuck Schumer and a group of, a bar partisan group of senators got together earlier in the year with sort of policy recommendations. It wasn't, it wasn't specific legislation. It was just sort of a set of policy recommendations. And that very clearly anchored towards competitiveness in American leadership as opposed to, for example, risk and safety issues, right? It was a very,

In fact, Tyler Cowan, I think, called the AI safety movement dead after that was released, which is obviously hyperbolic and for the sake of, you know, clicks on a blog. But still, it was, it was interesting how much it anchored to that. And now, you know, we sort of have a second and honestly sort of more important document in many ways that sort of, you know, reifies that as policy, basically. Yeah. Well, and I think what's interesting, too, is when you're looking at kind of U.S. leading from the frontier. There was, you know, a number of, you know, white space given to what U.S. allies and partners and how they could work together to play a leading role in a U.S.-led AI ecosystem.

So one of those examples is that the U.S. and the UAE struck a deal related to building AI-related data centers and energy infrastructure in the UAE. And it was criticized, but actually, this paper actually comes out to say that the U.S. network of allies and partners confer significant advantages over competitors and that the U.S. government must invest in and proactively enable the co-development of AI capabilities with select allies and partners, right? So it's not just, you know, being competitive at home, but being competitive on a global landscape. Yeah. Yeah. This is one that I'm going to certainly be watching more closely, you know, as time goes on.

But I want to shift gears a little bit, you know, like I said, we kind of bookended with sort of some government action around this. But, you know, a lot of the story of this month was also around the frontier model companies. And Open AI kicked it off with sort of the completion of their, this massive, you know, biggest crypto or biggest venture round in history. Yeah. So, you know, I guess maybe I'll ask it this way. given all of the controversy that they've had and the challenges, were you surprised that they were able to get this deal done at these terms? Or was this sort of just a foregone conclusion to you? I wouldn't say it was a foregone conclusion. I was not surprised that they got the deal done,

but, you know, $6.6 billion is a lot of money, right? That's a huge phrase. But equally, if you put it into context for a company that is as new as it is, this last week, the CFO basically came out with some numbers saying that 75% of its revenue comes from paying consumers and that they're converting somewhere between 5% to 6%

of their free users into paying users. So, you know, they are generating revenue, right? They are, there is value there, right? Isn't vaporware. If you look at who's invested in the round, I mean, some of those names, Microsoft and video makes sense, obviously that they have a vested interest in the company. But I think that the shift away from the nonprofit research, you know, start to open AI toward a for-profit corporation that is going to be reportable to shareholders.

And, you know, they have to invest in their profit-growing divisions where the majority of their staff already works, right? So, yeah, I think this was inevitable. It is an extraordinary amount of money. And it does beg the question, like, what are they going to use it for? Like, what are they going to develop next? And I think you and I were talking about. about, you know, chat GPT-5 and what's coming, right?

So, yeah, not surprised, but it is still a staggering amount of money. Yeah, I want to come back to GPT-5, maybe at the end of the show. Look, I think that your point is correct. One of the things, you know, at least their projections for revenue growth are absolutely enormous, right? They're projecting to get to, I think, 11 or 12 billion by the end of next year, and then something like 26 billion by the end of, you know, ridiculous. I mean, huge, huge amounts, which makes this valuation look not insane at all if they actually, if they actually achieve that.

Now, one of the really interesting wrinkles of this, I think, is the extent to which, you know, how much of that $6.6 billion is going to go to support and servicing their business that's growing versus racing to AGI. I mean, this is the really interesting wrinkle when it comes to frontier model competition is that there's an argument to be, made, and I think that this tension is on display inside these companies, that all of the business stuff is to some extent a distraction for the real battle they're fighting, which is for, you know, sort of getting to the most advanced models the most quick. So anyways, it's just, it's a fascinating thing, but, you know, I would anticipate, we haven't seen any sort of big follow-on announcements. There have been rumors of Anthropic raising at 40 billion and perplexity raising at 8 billion, but so far we haven't seen any of those come to come to bear yet. It's just, it's incredible. These numbers are incredible. I mean, some of it is just inflation, right? But yeah, it is incredible. I do think when you're looking at, as new models are coming online, you know, are they exponentially better?

Are they $6.6 billion better than its predecessor? Is chat 2505 going to be $6.6 billion better? You know, I don't think so. I think we're in this period of the innovation cycle where, you know, we're making incremental strides. Examples, a number of models came out over the last couple of weeks that have been trained in a way to be infinitely better at coding, for example. So reasoning, skills, et cetera. But I wouldn't say that it's like exponentially better, right? So I would assume for that level of money, they're going after. a bigger fish, right? This is not just about beating out Anthropic for retail market share.

Today's episode is brought to you by Plum. Want to use AI to automate your work but don't know where to start? Plum lets you create AI workflows by simply describing what you want. No coding or API keys required. Imagine typing out AI, analyze my Zoom meetings and send me your insights in Notion and watching it come to life before your eyes. Whether you're an operations leader, marketer, or even a non-technical founder, Plum gives you the power of AI without the technical hassle. Get instant access to top models like GPT40, Claude Sonnet 3.5, assembly AI, and many more. Don't let technology hold you back. Check out Use Plum, that's Plum with a B, for early access to the future of workflow automation. Today's episode is brought to you by Venice. Venice is a private, uncensored generative AI app. It

accesses open source models to enable text, image, and code generation without the fear of being spied on or having your data exploited. Discuss anything with Venice without concerns. about it being monitored, sold, or given to advertisers and governments. Venice is different because your conversations and creations are kept securely within the browser, never stored or accessible by Venice. Unlike other AI apps, Venice won't tell you what's okay to say or not. Venice won't patronize you. It simply provides direct access to machine intelligence, no topics are off limits, no ideas or taboo. With Venice, you're in control of the AI as you should be. Pro subscriptions are available for $49 a year or $8 per month. AI Daily Brief listeners receive a 20% discount on

Pro. Visit venice.a.ai slash NLW and enter the discount code NLW Daily Brief. That's NLW Daily Brief, all one word. Today's episode is brought to you by Super Intelligent. Every single business workflow and function is being remade and reimagined with artificial intelligence. There is a huge challenge, however, of going from the potential of AI to actually capturing that value. And that gap is what Superintelligent is dedicated to filling. Superintelligent accelerates AI adoption and engagement to help teams actually use AI to increase productivity and drive business value. An interactive AI use case registry gives your company full visibility into how people are using artificial intelligence right now. Pair that with capabilities building content in the form of

tutorials, learning paths, and a use case library. And super intelligent helps people inside your company show how they're getting value out of AI while providing resources for people to put that inspiration into action. The next three teams that sign up with 100 or more seats are going to to get free embedded consulting. That's a process by which our super intelligent team sits with your organization, figures out the specific use cases that matter most to you, and helps actually ensure support for adoption of those use cases to drive real value. Go to Bsuper.a.i to learn more about this AI enablement network. And now back to the show. Okay, so maybe to sort of again provide a little bit more logic from the investment side of this. We'll go to the next story,

which is not a story as a sort of a survey. The St. Louis Fed came out with a report that showed Gen. AI adoption happening faster than PCs or the Internet. Basically, they found that around 40% of American adults were using generative AI, you know, two years in or less than two years in, which is faster than any other cycle. I mean, again, sort of let's take the same questions. You know, surprise, not surprised, you know, validating or not like, you know, what was this, when you saw this report was it sort of just like, well, of course, you know,

That's what we all know. That's why we're living in it. No. I mean, I do think that I thought, yes, lots of people are playing with generative AI, right? And some of it being a tool and some of it being entertainment purposes. But the numbers, like 40 percent, that was, I was surprised by that, 18 to 64 year olds. And maybe I shouldn't be. So, you know, our users on Venice, the age range really does span.

I have one user in particular. I talked to him a couple times a week. And he's in his 80s. And he really just wants to learn and stay current. And he's lovely. And he's full of questions. So, you know, I know that there are users at both ends of those spectrums. I think what I was surprised by was how people are using it.

And some of the other stats were. of those adults that are employed, 28% use it at work. I thought that was interesting. That was higher than I would have thought, especially with, you know, corporate, you know, privacy and usage restrictions that a lot of companies place on AI. And that 33% used away from work. So that number didn't seem as odd to me.

And some of the tasks that people use them for are same across whether they're using them for work or home, like writing communications or, you know, performing administrative tasks. But, or coding, for example, is something that people would use it for at work. But some of the things that people are using it for at home, like their health and wellness and, you know, home improvement, recipes, like it really does seem to be kind of infiltrating people's everyday lives. a way that I didn't expect.

Yeah, it's interesting. We actually, we didn't put this on the list, but yesterday Apple finally started rolling out Apple intelligence. And one of the things that's fascinating is, I would say that the enfranchised crowd, the response all is some sort of like Steve Jobs never would have allowed this, you know, like this, this rollout is the worst thing that Apple's ever done. But if you just go cruise around Twitter, average users who don't care about all that and aren't paying attention, it's a lot of people.

saying it's really cool that it's really easy to remove unwanted objects from photos. Like there's these like the stupid simple use cases that people seem to be responding to. And that's Apple. My sense is that that's Apple's whole bet is that it's not going to be sort of big, complicated things. It's going to be little tiny, you know, everyday experiences that people find useful and just sort of get normalized. Well, that's their USB, right?

Removing the friction and making things feel and look as intuitive as possible. That is why the iPhone has been so successful. because it intuitively, when you're trying to navigate, it does things the way that you would expect that it would do, right? So there's a lot of kind of human intelligence put into how people use these technologies and devices. And I'm sure they'll apply that to AI. The one thing I would just say, obviously, from my own personal perspective, is a lot of the ways that people are using AI are incredibly helpful. But please everyone remember, if you're uploading a health report or your, you know, health tests, you know, the results of your MRI or something into, you know, an LLM or an AI system,

if there is no privacy feature there for you, that information will be used, either attached to your identity, and potentially used to know more about you, certainly used to inform the LLM itself. And, you know, want to encourage everybody to be thinking about those things as they start putting all this stuff into Apple Intelligence, which says it's going to be, you know, within a privacy, preserving infrastructure.

But, you know, if something exists outside of your browser or outside of your computer, then it could persist anywhere. So do keep that in mind. Warning noted. Okay, so let's talk about shift gears a little bit to a new sort of application that has been getting a ton of buzz. So there's not one particular story. But one of the big things throughout this month has been excitement around Google's Notebook LM,

specifically their audio overviews feature. You know, Andre Carpathy actually compared it to sort of, you know, he said it was the most fun that he'd been having with AI, you know, with an AI tool since ChatGPT was released, which obviously created a ton of buzz and momentum around it. Have you had a chance to play with Notebook LM yet? I have a little bit. Yeah. My co-founder, Eric, has also played with it.

And we are trying to use it to summarize all of our customer support tickets. So the idea is if we, obviously, we're able to anonymize those, put the whole support ticket from beginning to end of the conversation. and then turn it into basically, you know, a short podcast that the team can listen to quite quickly, you know, a five-minute, you know, audio file as opposed to like reading through a boring report to better understand our, you know, user experience, the themes that are coming through and our support tickets and the product feedback we get from our users. I wish I had actually done it in advance of today.

I should probably have a draft of it later. But yeah, we've been playing with it. We think it's really cool. Yeah, that's awesome. So what I think is interesting about that, that use case that you just described is my perception is that, or my belief, rather, is that people are very quickly going to find uses for this that don't have an easy one-to-one equivalent before, right? The idea of being able to take things and turn them into it, you know,

this tiny conversational summary, it's not going to be people. Sure, some people will create podcasts, like actual sort of like podcast unit podcasts with them. I think I saw someone put together, I actually maybe Carpathy himself did like a history of, you know, history's mysteries or something where you know, he put a bunch of interesting mysteries from history into it and created a podcast. But I think that the much more interesting things will be where people just hadn't used audio in this way before, right? So, you know, creating something. of tickets, like, that you can consume on the go, you know, or things like that. I think it's...

Well, it's a summary. So as I understand it, because, you know, we were discussing what the best way would be to, you know, populate the information, right? Like, what is the data set we wanted to use? And we didn't want an audio readout of what we already know, right? But somewhere in there over the course of, I don't know, 100 tickets, let's say, there will be themes, right? there will be people that are discussing, you know, the same types of issues, for example,

or we may be getting the same piece of feedback over and over. And again, that's something that we can understand already quantitatively. But within those conversations back and forth, there will be sentiment, right? Which isn't. It's more qualitative, right? And will the AI be able to kind of extract that and give us a. some more kind of thematic overview versus just the qualitative data. That's what we're looking for. Like, so like what's the what there? You know? Yeah. Super interesting. I love it. I'm,

I'm so excited to see another, another product that's just lighting people up. Meta actually just released an open source sort of build your own notebook LM type of thing. They call it, I think notebook Lama predictively. It came out yesterday. So, And listen, you know, it's got Google, you know, it's got people excited about Google again, which is, you know, been a curious, you know, non-feature of a lot of the last couple years of generative AI, you know, and I think that that product team is really stoked. And that's creating its own momentum, which I think is very cool. Well, maybe you'll have to try it one day. Put all of your show notes in and see if it can replace you. Yeah, we'll see. I, you know, the, I've often thought about. So we actually have on that line. So I've used it before as a sort of alternative to the long

reads episodes where I had it, you know, summarize and kind of do a conversational version of that. And it did a pretty good job. This was pre you being able to direct it. And one of the things that was noticeable to that, it was about the Longshoreman strike. And it definitely had a much more, I don't want to use sort of like the cliched woke term, but it definitely had a more, it really leaned into the human side of these issues and how much, you know, people need a stake in sort of, you know, the future in a way that I might not have if I was trying to summarize it. But now they've released updated features where you can help guide it. So that's really interesting. I've wondered for a while what the lifespan will be on the headlines section of the show, right? Because the podcast, the daily podcast normally has the headlines and then the main episode. The main episode is where there's a little bit more analysis. Now, I do do a fair bit of sort of like, you know, my own sort of, you know, narrative, you know, contextualization even for the headlines. But if there was going to be something that was going to be replaced, you know, that part that's just sort of summarizing the news, I think,

you know, I bet it could do a pretty good job. So speaking of Google, next story I wanted to touch on was Google and others. There has been a massive, massive shift, uh, towards nuclear as as power constraints come on. So, you know, this has been coming for a while, but this month alone we got Microsoft trying to restart Three Mile Island, Amazon signing agreements to do nuclear projects, and Google teaming up with Kairos on a nuclear power or energy arrangement. You know, what do you make of all this as it relates to, you know, well, one, I guess, you know, if we keep going back to this sort of surprise, are you so? Are you so. surprised to see how fast the sort of Overton window on this is shifting given how, you know,

on the outs nuclear has been for so long? Or is it sort of, is it maybe not shifting as much? And, you know, this is going to become an issue as the public realizes that, you know, all these big tech companies want to go nuclear. Yeah. I would, I think if I had not been as invested in Bitcoin mining policy issues as I had been over the last couple of years, it probably would have surprised me. But the shift. toward alternative energy sources for technology that is energy intensive is not new, right? So it doesn't surprise me from that perspective. You know, Amazon's based in Washington.

They're doing quite a lot with energy northwest up there. Washington State has a history of using quite a lot of nuclear. I think, again, I'm based in Europe where there's a lot of nuclear power and, you know, it has been quite safe. So the whole idea around, to me, I think it's crazy that we don't use more nuclear. So, no, I'm not surprised. I think it makes a lot of sense. And the fact that kind of everybody's doing it all at once, I think is really interesting, right? I think the goals to be, you know, transition to carbon-free energy are all kind of aligned with, oh, gosh, I've just done a blank on what the accord is the Paris Accord, I think, is the agreement.

And I think that those were, again, kind of 2030. So I think that Amazon's goal here was 2030 and Google's goal was 2035. So it would be interesting to see if they start, you know, competing with each other to actually bring their their carbon-free energy usage up and it becomes a differentiator. Yeah. You know, it's fascinating. So one at Davos or at the beginning of this year, Sam Altman got asked questions about energy quite frequently. And his response was basically always some version of. Yeah, right? You mean because it can be completely theoretical and known as a lot of. to do anything about it, but they feel important for having the right answer in their own mind. So his response, though, I thought was really interesting, which was basically we've never had a better financial incentive to solve energy issues. You know, like basically his base case for AI was that we have to solve energy issues

for AI to get to where it needs to go. So it's like he's not contemplating the AI future where it's just sort of used all our resources because it's just not possible, right? To get the AI where we, where we need it to be, we're going to have to do something. And this is sort of, you know, obviously he's, he's not just talking about bringing nuclear capacity online. He's thinking about, you know, other approaches as well. But, you know, we see a little bit of that coming to fruition, I think, with a set of announcements. Yeah. What I think will be interesting to see, and I don't want to kind of front run November. Maybe you'll have to have me back, depending on what the outcome of the election is. But

you know, will we see a massive shift in some of these policies and the U.S.-based companies' commitment to, you know, some of these activities that, you know, might sit a little bit more on, you know, the progressive side of the fence when, you know, I was watching a Trump rally last night and, you know, he's back to his drill, baby drill mantra, right? So I wonder how this is all going to look in three months' time if we have a change in the White House. Yeah. It would be fascinating to watch. Yeah.

Okay. So cruising through the last sort of few of our stories, AI won a bunch of Nobel prizes this one for chemistry and physics. Right? Yeah, like fringe. So I guess the question is, is every scientific discipline now just AI? Is it just AI everywhere? the way down? Well, you know, I think what we're seeing is how AI is going to permeate everything humanity does, right? So we are going to see increased efficiencies and really the incredible

reach of, you know, humanity's capacity. So the two nobels were for physics, you know, which makes sense on neural networks. So, you know, those two things kind of go together. But the other one was actually the chemistry prize. And it was for a, had something to do with proteins and predicting the structure of proteins, which was done through the use of some of the deep learning that Google's deep mind had pioneered. So yeah, I mean, it feels tangential, but I think you're right. I think it is a real nod toward all the things that may be possible.

Like if we're ever really going to go to Mars and live there, like, we're going to have to extend the capacity of human capability. And how do we do that? I think machine learning is going to increase it exponentially. One of the ways, you know, there's frequent moving targets on what the heck we mean when we say AGI, right? it kind of keeps shifting, but it also, you know, in part because as we get closer to it, it feels like the definition that we had before might not be quite right, but also because just I don't think anyone has ever had a really, really good, clean explanation. One of the explanations that I've seen some offer is sort of less a definition and more

a capability. It's like when AI can do to make actual scientific discoveries on its own is sort of the, the AGI moment. So I don't know, it's just sort of interesting to see in light of these Nobel Prize. as being, you know, related to AI in some way. Yeah. Okay. So the, obviously sort of a lurking theme behind many, many conversations right now is the agentic era.

And, you know, how quickly we move from sort of just the assistant era of AI, of generative AI into the agentic era. And we had a really interesting moment along that path this month when Anthropic announced their computer use. So basically, you know, the quad can now in limited circumstances via the API, you know, click, click, you know, on screen and do things because of that. It's a really fascinating way that they got it to do that. It's, you know, measuring pixels between different areas. But, you know, what do you think?

Like, how significant is this announcement? Is this announcement a big one in and of itself? Is it a big one in terms of what it represents going forward? You know, what did you think when you saw this? Yeah. I mean, we also had something happen this month where there was an AI bot that started a crypto meme or something. Yeah, we could have gotten deep on that one. I mean, actually, going back to what I was just saying, another new Turing test that someone had tried to make, you know, a thing a few months ago was the first time AI, you know, makes a million bucks, which I think technically this would count as.

Yeah. Well, depends. Is it paper money? Is it a real money? Yeah. But yes, I think that obviously the things that we can get, that is the whole point of the whole discussion around AI agents, right? And they're not, they're sovereign, but they're not, you know, AGI. Right.

They're not necessarily thinking. They're taking instruction and they're able to do something unassisted, right? So it does seem like the natural next step. And ultimately, having something perform a task for you. as an assistant, so you're still quite involved in the production and the outcome, as opposed to just saying, this is the outcome that I want, go and do that. I think is one of those things that is going to kind of leapfrog forward, not just adoption and usage, but really what the value proposition is here.

And so, you know, maybe that's going back to our first discussion. Maybe that's what Open AI spends that $6.6 billion on. Yeah. So maybe I'll weave on our last conversation into this. So the last conversation is not news, but rumors, right? So there were rumors that came out. The Verge started reporting that GPT5, which is, you know, or what they codename Orion internal to OpenAI was coming out by December.

We've also got reports that Google's Gemini 2.0 might be. for December. What do you make of all this? Do you think that the battle is around sort of the next big frontier model, or do you think it's going to be sort of more about agent capabilities? That's a really tough question. I mean, I don't think that there's going to be, how do I want up with this? I think we're always going to be evolving the models, right? The models are still, you can train them and fine-tune them to be very specific. to do certain tasks better than others, but we still don't have like one model to end the models that can do all of the things, right? And going back to kind of human behavior, we like less friction.

We like to be able to just kind of put everything into one place and then have it kind of do stuff for us. So I don't think that's going to stop because I don't think that we've gotten to, you know, the model of all models yet. However, again, going back to usefulness, there needs to be more use cases, I think, other than, I mean, to get me wrong, I use AI for search now. I very rarely search. Obviously, use Venice quite a lot for that, 99% of the time. But, you know, another thing that I was working on this last week had to do with understanding the payments that come into the business that I run, right? And the reporting mechanism with my payment service provider was just woeful. It was not able, despite nearly a week of me looking at every report, to answer the

questions in one dashboard that I had. And then it became, well, do I dump a CSV file and go through an SQL process? And, you know, finally, I just kind of put this question into an LLM and said, I run a business, it does this. These are the numbers that I have. Build me a financial model. And it did. Now, I need to know enough about my business to understand if it's on the right track to be able to kind of tune that. But, you know, I could hire a quant to build this financial model or, you know, I can kind of prompt the LLM to do this for me. It would be great if we got to a point where I could just say to an agent, I need this information summarized for me in a actionable format. And this is all I'm going to tell you. And it would just go and it would do

it and it would understand that this is new information that needs to be brought in or this is information that isn't relevant anymore and be able to kind of undertake those tasks. And I think that's where agents come in. And I think that will be really, really useful. That will change the way we work and the way we live. Yeah, I agree. I think we're at the point where, you know, agents have been the next big thing since basically the moment that ChatGBTGBT launched and people started thinking about it. We are finally starting to see some of these experiences come online. I think it's it's probably inevitable that there are going to be real limits to exactly what they can do for for some time. But boy, is there a lot of entrepreneurial energy around

trying to solve that set of challenges. Well, and it gives crypto something to actually do, right? Because these agents aren't going to be able to transact with, you know, their Citibank bank bank accounts. So I think that's really exciting. That's a whole other topic. Yeah, considering I can't even interact with my Bank of America account because of crypto. Well, Tiana, awesome to have you on the show. Really, really interesting thoughts. It felt like a consequential month, but I think, you know, there's a significant possibility that next month makes it look quiet by comparison.

Yeah, we'll have to wait and see. Thank you again, though, for spending some time with us today. Thanks for having me. It was great to be here.