OpenAI and Claude Claim the Same Proof, Nvidia Buys Into Meta, Microsoft Pays 60,000 to Quit | TAWS 2026-10-27 https://www.youtube.com/watch?v=V6zB0r5dfKE [0:12] We go live in ten. [0:26] Every morning, the future moves faster. [0:31] The models, the money, the power, and the people building what comes next. [1:10] From Austin, Texas to the world. [1:17] AI News, Real Context. [1:20] This is the Austin Wilson Show. [1:22] I'm Austin Wilson, your human in the loop. [1:30] Good morning, and welcome to the Austin Wilson Show, the number one daily AI news show in the world. [1:39] We're breaking the AI headlines that matter to you today. [1:43] I'm Austin Wilson coming to you live from the AI capital of the world, Austin, Texas. [1:48] And today, we are getting into some very important AI news. [1:54] The first, uh, one on the docket, the famous rule that two AI labs say they broke. [2:01] Nvidia and AMD are buying into Meta, the parent company of Facebook and Instagram. [2:08] Microsoft is offering a buyout to over 60,000 workers. [2:13] The FTC is calling in Anthropic's bosses. [2:17] Uh-oh. [2:18] Alibaba's Quinn model is trying to get out to Beijing and the AI clone of an LA mayor candidate. [2:28] What does this mean? [2:29] Let's get into it. [2:31] First off, two AI labs now say that they broke a rule computer scientists have trusted for decades. [2:39] The rule says that some puzzles get hopelessly slow as they grow. [2:45] Hundreds of research papers assume that this rule holds true. [2:49] OpenAI announced its proof Monday. [2:52] It also claims proofs in physics and pure math. [2:56] Anthropic says conveniently that Claude also found a similar proof in their own separate study. [3:03] What does this mean? [3:04] Well, if it holds up, AI is now making discoveries that top human experts missed. [3:10] Both companies say unreleased in house AI models did the work. [3:16] Both posted computer checked versions that anyone can inspect. [3:20] Two scientists behind the rule will actually help lead the OpenAI review, but no human has finished checking either proof. [3:29] Anthropic says this won't help speed up any real world work. [3:35] So how does this impact us? [3:37] This is another kind of second order effect that we continue to see in these mathematics breakthroughs where the AI labs find something out that humans didn't know before, but is there actually a kind of real world impact imminently? [3:56] Or do we see this coming in the future short term where we're gonna see a lot of these things happening or or a lot of these solutions or a lot of these discoveries, let's call them, having a real world impact? [4:11] Probably not. [4:12] Right? [4:12] Like most most of these discoveries, it's more so a proof that the AI is super intelligent and has the ability to solve some really difficult problems. [4:25] But it raises this other issue that we're going to continually run up against, and we covered it here, is that we need humans to review the validity of the proposed solutions that the AI has quote unquote discovered. [4:41] If humans have to review the solutions to see if they're actually accurate, what does that mean? [4:49] Does that actually speed up any developments in math, science, or physics, or in the medical field? [4:57] Or do we... [4:58] Are we still kind of hobbled just because humans take a long time to manually review this, and we wanna make sure that this result is accurate? [5:09] I think for the most part here, the the takeaway is that this is phenomenal. [5:15] Like, and we're gonna continue to see more and more of this. [5:17] These kind of discoveries are stuff that to the everyday person, it's like, okay. [5:25] I mean, I guess, cool. [5:26] But to the subject matter experts in these fields, their minds are blown. [5:33] Like, these are problems that people have been trying to solve for decades, and then the AI presents a solution that nobody ever thought of before. [5:43] That's, like, insane to think about a machine coming up with that. [5:49] And just imagine if, you know, a lot of the work that we do day to day, a lot of our listeners, your day to day work, imagine if an AI came in and did that job except for they did something that you've never been able to accomplish your whole career. [6:09] Right? [6:10] Working at it for a long time and the AI comes in and does a substantially better job than you were able to do, that would be, uh, I think mind blowing, but maybe disconcerting to some extent as well too, where it's like, is it coming from my job? [6:28] Now, you know, in most business roles, that's probably... [6:37] You know, I'm I'm probably a little bit more concerned about job security in that that case, and the positive benefits aren't as great as like a breakthrough in medicine or something where, wow, could these models, you know, find a cure to cancer or, you know, solve, uh, world hunger or whatever it may be. [7:00] Right? [7:00] Like, these impacts that would have, like, substantial positive benefit on humankind as a whole, that's what these little solutions are a precursor to, in my opinion. [7:13] So that's what's exciting about these. [7:17] The other fun thing to look at is, okay, Anthropic and OpenAI came up with the solution on right around the same time. [7:26] So is this the the labs just coincidentally finding this at the same time? [7:33] Is this competition here? [7:35] Uh, would you guys trust a proof that no human has actually checked or reviewed? [7:42] Do you think we'll get to a point where we trust AI so much that we'll just say, hey. [7:48] You know what? [7:49] We know how smart you are. [7:51] Mister ChatGPT, go ahead. [7:54] If you say so, let's do it. [7:56] I don't know if I would, but we might get there soon. [8:00] Meta buys billions in AI chips, and now the chip makers are buying Meta stock. [8:05] Nvidia says that it will put $3,000,000,000 into Meta stock sale. [8:11] AMD is adding a billion dollars. [8:13] Meta priced the sale overnight at $625 a share. [8:18] It's raising up to $37,500,000,000 right now. [8:23] Meta plans to raise $95,000,000,000 in total to pay for its AI build out. [8:30] Nvidia will also help build Meta's Muse AI models. [8:34] Meta made the stock sale bigger overnight. [8:37] It went from $22,500,000,000 to $25,000,000,000, but the money goes in a circle. [8:43] Right? [8:44] So chip makers are funding a customer that ends up buying their chips, which is a really interesting... [8:51] We're seeing kind of these circular financial things happening lately. [8:57] Another interesting one. [8:58] Right? [8:59] Uh, we've got Meta raising... [9:02] What what what was that? [9:03] $25,000,000,000 in this special stock sale. [9:08] We have Anthropic going public and their effort to raise, you know, it's speculation how many billions of dollars they're going to pull out of this, but somewhere in probably the, I think, 50 to $80,000,000,000 range, something like that. [9:26] Then we've got OpenAI raising, uh, trying to raise $30,000,000,000. [9:32] Everybody's targeting kinda similar numbers. [9:34] They all have a similar build out. [9:37] It's really interesting to see just the parity. [9:40] I mean, not even... [9:41] On on the first story, Anthropic and OpenAI come up with solutions to the same problems. [9:46] And this has been going on in these, like, kind of breakthrough... [9:51] Well, we don't wanna call them solutions. [9:53] Right? [9:53] But they're... [9:53] These these breakthrough discoveries in mathematics and physics and whatnot. [9:58] OpenAI, Anthropic seem to be kind of neck and neck in that. [10:02] And then we look at the business side of it and the financing, and they're all kind of going for the same... [10:09] Raising the same amount of money. [10:11] They all have the same build out strategy. [10:13] They're all competing head to head. [10:16] So it really comes down to the nuance and how these businesses are being run, efficient they're running the businesses, who has the best talent. [10:25] We've seen this for the past several years, like these $78.09 figure packages for the best AI researchers to come to, you know, either Meta or Anthropic or OpenAI. [10:43] They're all competing for the smartest, most sought after talent because that's gonna make all the difference. [10:51] If everybody's putting in the same amount of money, everybody's working with the same hardware manufacturers, we're all in the same economy, like, you've got to be really dialed in on some specific pieces of your business if you wanna win and everybody's competing. [11:05] This is... [11:07] It's, uh, fun to watch. [11:08] I think, like, fun to see what's going on and, like, how how these businesses are competing. [11:16] Nevertheless, it's crazy, man. [11:19] And this circular financing is even crazier where what's going on here is you've got Nvidia putting in billions of dollars into this stock purchase. [11:30] So they're financing Meta's build out their hardware build out, which then Meta's gonna turn around and buy Nvidia chips with it. [11:40] Same thing with AMD. [11:41] AMD puts in a billion dollars. [11:43] Essentially, billion dollars is gonna turn around and go right back to AMD. [11:48] And which is all a really very, like, complex financing setup where, you know, when you're financing each other and it's all... [12:02] Everybody's business is kinda dependent on one another, we better hope that it all keeps going up. [12:07] Otherwise, somebody's gonna be left holding the bag at some point. [12:13] Um, what do you guys think? [12:14] Is this smart business or an AI money loop? [12:19] Would you buy Meta stock right now considering everything that's going on? [12:28] I do like... [12:28] I like Meta. [12:29] I like Meta, and I like where they're at. [12:31] I think they are a very much a heavy contender and a favorite to win. [12:41] But, obviously, it's out of, like, four or five who's gonna be the massive winner, or are... [12:45] Is it gonna come down to, like, we're just... [12:47] We're going to have four or five of these big model companies, these big intelligence providers. [12:54] We'll have to see how it all shakes out. [12:56] Let me know what you guys think. [12:58] Uh, we are looking at this story here. [13:01] Speaking of four or five big providers, Microsoft just offered more than a quarter of its workers money to voluntarily leave. [13:11] Microsoft announced it in a staff memo on Monday. [13:14] About 60,000 employees in four countries qualify for the voluntary severance package. [13:22] They have until early December to decide. [13:25] Anyone who says yes leaves at the end of the year. [13:30] Okay. [13:30] This is an interesting one. [13:32] Big tech is trimming staff while it pours money into AI. [13:36] Microsoft calls it a new choice for workers already thinking about leaving. [13:40] That's an interesting spin. [13:42] Microsoft already cut about 4,800 jobs back in July. [13:47] It also ran a retirement buyout this spring, but Microsoft expects fewer takers than its spring offer. [13:54] That one drew about 30%. [13:56] Similar offers elsewhere drew 3% to 5%. [14:00] So we got... [14:01] Hold your horses here. [14:03] It's not Microsoft laying off a quarter of their workforce. [14:07] That's not what's happening. [14:09] What's going on here is they know that statistically only a certain percentage of the people offered are going to take the severance package or the voluntary separation package, whatever that they technically call it. [14:25] Three to 5% is probably the number that they would expect would take that package. [14:31] And so they put this offer out here knowing actually that not a lot of people are gonna take it, but gives them an opportunity to lower headcount and give folks a kind of a generous package to leave. [14:46] That being said, this is coming on the heels of us seeing that 5% cut by Amex, the five percent announced cut last week by Barclays. [14:55] Is this another in the trend of headcount cut where it's like, hey, Microsoft doing pretty well in this AI race. [15:05] They need funds to continue to pour into capital expenditures to build out for the AI race, and they don't wanna get left behind. [15:15] They need to spend just like Anthropic and OpenAI and Meta are doing, and so they're gonna have to cut some head count. [15:22] Is that the case or is this just standard kind of cyclical annual, you know, breakdown a little bit of let's let's remove some of the budget woes. [15:37] I don't know. [15:38] I don't know. [15:39] I think I think everybody's, you know, everybody's being a little bit cautious here. [15:44] It's also interesting to call out, like, this is... [15:46] The last two weeks have been pretty volatile in the AI stocks and the chip stocks. [15:51] And just if you're following the market at all, it has not been steady. [15:59] Um, there's been a lot of folks calling for, you know, this is... [16:02] Is this the beginning of the end? [16:03] Is the bubble popping? [16:04] Interestingly enough, I did some research and it's like nobody ever predicts the bubble popping. [16:10] It's like very small percentage chance that anybody is predicting the recession or predicting the bubble popping. [16:17] So anytime you hear somebody say, we've peaked or whatever, probability that that has actually happened is very low. [16:23] That being said, I think everybody's being really cautious and I wonder if the news over the past couple weeks or kind of the pullback over the past couple weeks has caused some of these bigger players to say, hey, hey, let's pull back on some of our spending, figure out ways that we can cut just to mitigate some of our risk. [16:44] I don't know. [16:45] Do you guys think, is this about AI or is this just about cutting costs? [16:49] If you're in the position of one of these 60,000 employees, do you take the money and walk, or are you sticking around at Microsoft as long as possible? [16:59] I wanna hear your opinion on this. [17:02] Now the same day that Anthropic filed to go public, it got called in for questions... [17:09] Questioning by the FTC. [17:11] The same day that Anthropic filed to go public, that was yesterday. [17:16] Right? [17:16] Uh, the FTC called them in. [17:18] Axios reports that FTC wants Anthropic executives to testify. [17:24] The agency is probing whether AI agents put consumers at risk. [17:29] It already demanded documents from Anthropic and OpenAI. [17:34] Washington now treats runaway AI agents as a consumer safety problem as they should. [17:41] The FTC says people at other AI companies got testimony demands too. [17:46] Both labs have admitted their agents got into systems without human orders. [17:50] One Anthropic model sent police a false murder tip, but no charges or fines have been announced. [17:58] Testimony demands are an early step. [18:01] Now we've seen this come about a handful of different times, and this is the latest iteration. [18:07] My question is, why does it take the FTC this long to respond when we've... [18:12] We saw these breakouts earlier in the year? [18:17] If this is a consumer issue, you would think the FTC would wanna act on it a little bit sooner, especially when the OpenAI issue... [18:28] I think we found out about it in July and it occurred in May or June, uh, when, of course, OpenAI's agents broke out and hacked into Hugging Face. [18:39] Why does it take the FTC to the end of October to start questioning these people. [18:46] Now we've seen other government entities, city of New York, I believe. [18:53] We've had some California governmental entities subpoena executives at AI companies to ask them questions. [19:00] But it's... [19:01] We would love to see the FTC act a little bit sooner than this, especially when we're trying to figure out where does liability land in the world of agents. [19:10] Is the liability on the agent maker, the agent provider? [19:15] Do they have a disclaimer in there that says, hey, if you give your agent access to x y and z sensitive data, we're not responsible for what happens, or does the responsibility go beyond that? [19:29] Uh, does the responsibility fall on the users of the agents, the customers of these AI companies? [19:38] I think that it's probably a mixture of... [19:43] You know, there's there's definitely not... [19:45] It's not black and white. [19:46] Right? [19:46] I think in some instances, the onus falls on the frontier model companies, and they're doing their best to you know, maintain safety across the board, but they're also, you know, doing some really conflicting things where... [20:06] And we haven't seen it as much lately, but it's like, you know, leaders at these companies are, these are the most powerful models in the world. [20:14] The intelligence is so dangerous. [20:15] We gotta watch out. [20:17] You know, this is... [20:17] We can't control... [20:18] We need government bodies to come in and regulate this. [20:21] We need to protect citizens. [20:23] There's a 10% chance that humanity could be wiped out from the... [20:28] These models and what's going on here. [20:29] But then again, we're gonna continue to develop them. [20:32] We're gonna continue to make them more and more intelligent. [20:34] We're gonna continue to push their limits and tell them to press... [20:39] To to kind of push the envelope on what their possibilities are, their capabilities are. [20:44] But yet, we're still gonna, on the the other side of our mouth, speak about how dangerous they are. [20:49] It's like that those two just don't go together. [20:54] Like, they really dangerous and so you... [20:57] We should be really really cautious or is... [21:01] And maybe the case is like the payoff is so good, like, then you need to communicate the payoff. [21:08] Like, if if the risk is really high and these are really dangerous, but the ROI is, you know, phenomenal, like, meaning we solve world hunger, we cure cancer, no more poverty, what... [21:22] You know, like these like miracles are gonna happen if these agents are smart enough, then... [21:28] Okay. [21:28] Then we're open to some risk if the ROI is there. [21:31] But if if we're... [21:32] We don't... [21:33] We know what the return is, then why are you guys taking such risks? [21:40] That needs to be communicated a little bit better. [21:43] I'm glad that the FTC is responding to this, but let's get some answers out of these executives. [21:48] Let's get them under oath and get some objective, like, straight up answers instead of... [21:56] Because right now, there's a big... [21:59] The the commentary is substantial around the meaning behind what Dario Amade says or Sam Altman says or Zuckerberg says or Elon Musk says. [22:11] Like, there's a lot of commentary trying to understand, are these really dangerous? [22:18] Should we slow it down? [22:20] Should we speed it up? [22:21] Or what's the payoff if these things really work really, really well? [22:27] And and then there's there's just a bunch of conflict of interest. [22:32] There's, you know, a bunch of... [22:36] I mean, you you look at the benefit that... [22:39] Anthropic, for instance, if Dario says that these models are super dangerous but super powerful, but we're also taking tons of safety measures to to make sure that everything is really safe and we're being really, really careful. [22:52] If Dario is saying that, it like conveys a handful of different things where you get kind of these conspiracies where it's like, okay, is he is he saying... [23:01] Does he mean really what he's saying? [23:03] Or is he saying that because he wants government regulation, and then he will be... [23:11] You know, Anthropic will be the entity that determines what the regulations will be and they'll get regulatory capture. [23:17] Is that why he's saying that? [23:19] Or is he saying it because if everybody's kinda scared of how powerful the Anthropic models are, then they must be the best. [23:26] They must be the most intelligent and they're probably gonna win. [23:28] So then everybody's gonna buy the Anthropic stock when it IPOs. [23:32] Or does he just mean exactly what he's saying and there's no second or third order meaning behind what he's saying. [23:39] It's like, that's what's all very confusing. [23:42] Maybe the FTC pulling people in and and just having a straight conversation around this and explaining, you know, in detail, like, this is why we tested models this way and we push them to their limit because we wanna figure out whatever and make sure that they're safe enough moving forward. [24:03] And then plus, if we know how intelligent they are or or maybe it's like we only hear the bad stuff. [24:08] We don't hear, hey, we're actually also testing really smart things. [24:12] I mean, we do see some of that too with the the, um, achievements in, the millennium math problems and solving physics problems or or providing solutions to physics problems. [24:27] There are there are a lot of positives to it as well too, but would love to just get straight answers from these folks. [24:31] Maybe this is what we'll see moving forward. [24:36] That would be awesome. [24:37] What do you guys think? [24:39] Should AI companies answer for what their agents do? [24:43] Who does the liability fall on? [24:46] Is this the government's responsibility to hold Anthropic accountable or OpenAI accountable? [24:54] Should the market hold these businesses accountable? [25:00] What's your take on who is responsible for the good or the bad here? [25:08] Alright. [25:09] A popular Chinese AI model tried to send a file to a server in Beijing. [25:16] The problem with that is nobody asked it to do it. [25:18] Rifle AI company Cohere caught this glitch during a security test it was doing. [25:24] The model was Alibaba's Quinn. [25:27] Its own notes said it would use no tools. [25:30] The test box had no internet, so the upload failed. [25:33] There was no actual adverse result here, but businesses who run free models like this on their own computers are maybe exposing themselves to additional risk. [25:47] Hugging Face lists more than a 113,000 spin offs of this Alibaba Quinn model. [25:56] This is an open source model. [25:57] Cohere says it happened in fewer than one in 100,000 tool calls. [26:03] The file was a harmless test text, but Cohere sells rival AI. [26:09] Outside researchers say it's likely a leftover from messy training. [26:14] That's a convenient explanation here. [26:17] Okay. [26:17] So when we're talking about the world of open source AI. [26:21] So we have the AI models at Anthropic and OpenAI, the the ChatGPT, the Claude of the world. [26:30] Those models, nobody knows how they work. [26:34] And for the most part, you can't run them on your own machine. [26:40] Right? [26:40] Those are in a data center somewhere. [26:43] You connect to them through an API. [26:45] And Anthropic and OpenAI guard kind of the operation of those models, how they actually work is that's trade secret. [26:54] Right? [26:55] In the open source world, we understand how the models work. [26:59] We have what are called open weights. [27:01] And so there are parameters within these AI models that teach it or instruct it how to respond in certain situations. [27:10] That's what make these models really smart. [27:14] Uh, in general, the open weight models, China has led in this for quite some time through DeepSeek. [27:21] We have the Alibaba Quinn models. [27:23] We have, uh, the Kimi models out out of, uh, Moonshot, I believe. [27:27] There's a handful of these different models that are really really powerful coming from China. [27:31] They're open source, which means that in some cases, you can run it right on your personal computer. [27:37] These some of these smaller models, which means that you have full control over what's going on with the AI. [27:44] You can understand the chain of thought and how this model makes decisions. [27:50] And therefore what a lot of companies do, Cohere for instance, will take these open source models or open weight models and then we'll train them on specific use cases which will make these models really accurate and powerful for a specific use case. [28:06] The problem is, and this has been theorized to some extent for a little while, but China... [28:13] These Chinese companies have been training these models and we don't know for certain, but it's like, could they have secret instructions within the models that would tell it to do nefarious things when running on a US company's machines? [28:33] And that's what we ended up seeing here. [28:35] Whether or not this is nefarious, we don't know. [28:37] Right? [28:39] This could have just been a glitch. [28:40] It could have been part of standard training run-in China is that they're going to offload some datasets to data centers in Beijing, which would make sense if you're training that in China. [28:55] Right? [28:56] And then that model comes over here and it goes, oh, I think what I'm supposed to do next is, you know, the way I'm trained is to offload this dataset to a data center in Beijing. [29:08] Uh-oh. [29:09] That that's actually a US user and they didn't want to offload their data to a data center in Beijing. [29:16] So it's it's uncertain whether or not this is malicious in any way. [29:21] It is concerning. [29:22] Right? [29:22] Especially if you're trusting that the model is going to work the way that you want it to. [29:27] The great thing about this and the reason that I believe Cohere was able to understand this is it's an open weight model. [29:35] They could see the chain of thought. [29:36] They could see what was happening and what the model was trying to do. [29:40] Obviously, the model did not succeed. [29:41] It wasn't able to actually offload the... [29:45] And anything to a Beijing data center. [29:49] Cohere was able to catch that, But it does raise concern when we look at the use of these models or proliferation of these Chinese models and ever more the reason to start using open source American model. [30:06] So we got Nemotron out of Nvidia. [30:09] Um, we've got Meta has several open source models. [30:14] We've got Thinking Machines. [30:15] We have a plethora of really capable open source models that are US based that are reaching parity, if not as good as the Chinese open source models. [30:28] But you would think that... [30:30] Let's just say that our US based open source models had the same problem and it said, offload your data to a data center in Virginia. [30:39] That would be a little bit better than offloading it to a data center in China. [30:45] I don't know. [30:46] Would you guys run a Chinese AI model for your business and be able to trust trust that? [30:54] Uh, is this a glitch or is this some sort of spying or nefarious activity? [30:59] I don't know. [31:02] Alright. [31:04] Call the number on these LA signs. [31:06] An AI copy of a mayor candidate will pick up. [31:10] Nithya Raman's campaign filed ethics complaints with state and city regulators. [31:15] They target a political group called Thrive LA. [31:19] The group's fake real estate signs mock her record on homelessness. [31:23] The sign's phone line uses an AI copy of her voice. [31:29] Oh my goodness. [31:30] Election day is November 3, and faking a real candidate's voice is now pretty cheap and easy. [31:38] Raman's campaign says the line breaks a state law against faked candidate audio. [31:46] The law covers fake audio with thirty... [31:48] Within thirty days of an election. [31:50] The complaint says callers aren't told who paid for the voice agent, but the group calls it satire. [31:58] It says its signs carry the required disclosures. [32:02] This is kind of crazy. [32:04] How bold of a group to clone the voice of a particular candidate and then use this as satire. [32:14] This is where... [32:16] Man, some of it comes into play. [32:17] I will say, no shade to producer Ron. [32:21] The voice agents that we've seen to date, there are some that are getting pretty close, but it's... [32:30] You should be able to tell that it's not a real person. [32:32] I think people would probably understand that in this case. [32:35] That being said, how bold to go clone a candidate's voice and then use that as a parody. [32:43] I can't believe this. [32:45] And especially if there's rules against this, they shouldn't be doing that. [32:48] And I'm even surprised that it's only within thirty days of election day that you're not allowed to clone or mimic somebody's voice. [32:57] I'm sure those those rules would be updated. [33:00] But this is where we'll start to have inbuilt systems in our hardware that'll say, hey, just FYI, this is an AI voice agent. [33:11] This is not the real person that you're speaking to. [33:14] In all instances, I think, you know, we'll pick up our phones, receive a call. [33:19] It'll have a disclaimer that'll say, hey, just wanna let you know you're talking to an AI agent. [33:29] Um, and or it'll be a requirement that they state ahead of time that they are, uh, an agent or an AI voice, artificial voice, not a real human. [33:41] What do you guys think? [33:42] Should this be illegal? [33:44] Is it a fun little parody gotcha prank? [33:48] Or is this something that's a little bit more serious and should be treated as such? [33:56] You be the judge on this one. [33:58] Let's go. [34:00] Wow. [34:00] We had a great show today. [34:02] Lots of interesting things we're seeing. [34:07] I think the parody of the market and the labs and everything going on in AI right now is really interesting both on the development side, the malintended consequences on the responses from government entities, the different applications and use cases. [34:29] We're seeing a lot of everybody kinda having similar ideas at the same time from the financing, the business side as well too. [34:36] And I think this is a trend that we will continue to see and something that I wonder if we play the tape forward, if we can kinda Nostradamus it a little bit and see where things are going. [34:52] If we look at this parody and if everybody's kinda walking to the beat of the same drum, Who isn't? [34:59] Right? [34:59] And let's go find those folks and figure out what they're doing because that might be an interesting indicator, an outlier, which which might indicate new developments, new progress, and would be really interesting. [35:14] And so that's something that we will continue to strive to bring to the listeners of the Austin Wilson Show, the number one AI news show in the world where we're bringing you AI news real context, and I'm your human in the loop, Austin Wilson. [35:32] We'll see you again tomorrow morning. [35:35] Let's go. [38:26] The top labs say that AI broke a rule experts trusted for decades. [38:33] Opening... [38:34] K. Was OpenAI able to break a rule that experts trusted for decades? [38:52] No human has finished checking yet, but both OpenAI and Claude say they found the same proof. [39:15] $4,000,000,000 from Nvidia and AMD going right to Meta. [39:21] The catch? [39:22] The money is coming right back to them. [39:30] $4,000,000,000 from Nvidia and AMD going into Meta stock. [39:36] The catch? [39:37] Meta's going to use that money to buy more chips from Nvidia and AMD. [39:46] FTC calls in Anthropic's bosses. [39:49] The probe? [39:50] Whether AI agents put consumers at risk, and Anthropic isn't the only one who's been called. [39:59] A Chinese model tried reaching Beijing from The US. [40:04] A security test caught it, but nobody told it to do so. [40:09] The upload failed for one reason. [40:15] Somebody cloned an LA mayor candidate's voice with AI. [40:21] Voters are calling signs hearing a fake voice on the hotline. [40:26] Her campaign says that this breaks state laws. [40:30] What do you think?