The Matt Walsh Show - September 10, 2026


Ep. 1837 - Only A Few People Win from AI Hysteria. Everyone Else Loses.


Episode Stats


Length

45 minutes

Words per minute

165.18

Word count

7,469

Sentence count

433

Harmful content

Misogyny

4

sentences flagged

Toxicity

4

sentences flagged

Hate speech

10

sentences flagged


Transcript

Transcript generated with Whisper (turbo).
Misogyny classifications generated with MilaNLProc/bert-base-uncased-ear-misogyny .
Toxicity classifications generated with s-nlp/roberta_toxicity_classifier .
Hate speech classifications generated with facebook/roberta-hate-speech-dynabench-r4-target .
00:00:00.000 Everybody has that list. The things you know you should get around to eventually.
00:00:05.060 Organizing the garage. Updating important documents. Taking care of long-term financial
00:00:10.500 planning. Life insurance tends to live on that list for a lot of people. Not because it isn't
00:00:15.900 important. Because people assume it's going to be complicated. That's why Policy Genius exists.
00:00:22.720 Policy Genius is an online insurance marketplace that helps you compare life insurance quotes
00:00:27.880 from top insurers in one place, for free.
00:00:31.540 Their license team helps explain your options clearly,
00:00:34.680 compare coverage amounts and prices,
00:00:36.940 answer questions, and help you find a policy
00:00:39.280 that fits your needs.
00:00:40.780 No guesswork, no confusion, just clarity.
00:00:44.660 And while they're helping with the shopping,
00:00:46.680 you can get back to enjoying summer
00:00:48.560 with the people who matter most.
00:00:50.880 Policy Genius has helped thousands of families
00:00:53.100 find the right coverage
00:00:54.300 and has earned thousands of five-star reviews
00:00:56.780 in the process.
00:00:57.880 because some items on the to-do list are worth checking off.
00:01:01.480 With PolicyGenius, you can see if you can find 20-year life insurance policies
00:01:05.660 starting at just $276 a year for $1 million in coverage.
00:01:11.320 Head to PolicyGenius.com to compare life insurance quotes from top companies
00:01:16.380 and see how much you could save.
00:01:19.000 That's PolicyGenius.com.
00:01:21.800 Opinions on AI seem to be, to put it mildly, sharply divided.
00:01:26.660 But one thing we should all be able to agree on, a statement that should attract a wide consensus, is that this is an incredibly consequential technology, advancing and scaling at enormous speeds.
00:01:39.020 And both the benefits it provides and the threat it potentially poses should be seriously considered in an objective, sober-minded way.
00:01:47.340 If we can agree on nothing else when it comes to AI or anything else, we should be able to agree on that.
00:01:54.140 It's not a political statement, it's not an ideological statement, and yet sober-minded objectivity is the very last thing we're seeing on this subject from either side of the debate, and the events of this week perfectly illustrate that point.
00:02:08.500 So let's start with those events, which immediately reveal why sober minds are not prevailing and why they desperately need to.
00:02:18.320 There is a colossal amount of money and power at stake.
00:02:22.180 For example, let's say that you're a massive, privately held artificial intelligence company
00:02:27.260 and you're planning a blockbuster IPO in a matter of months, which will probably value
00:02:32.380 your company at several trillion dollars.
00:02:35.580 You have every incentive to promote your product and to convince as many people as possible
00:02:40.800 to sign up and pay a subscription fee.
00:02:43.580 Why in this scenario would you go on national television and tell everybody that your product
00:02:49.100 has the potential to eliminate all of humanity within the next decade. Why would you deliberately
00:02:54.320 torpedo public confidence in the entire AI industry at a pivotal moment for your company?
00:03:02.180 You know, it seems a bit like Taco Bell announcing a new burrito by explaining in great detail how
00:03:06.720 it's going to, you know, absolutely demolish your intestines and potentially all of the
00:03:11.120 intestines on earth. Doesn't seem logical. Seems like self-sabotage. And yet that's exactly what
00:03:17.620 Anthropic, which makes the AI model known as Claude, has apparently just done, providing an
00:03:23.680 incredible window into how the most powerful corporations on the planet can manipulate public
00:03:28.620 opinion and the variety of tactics they use to push more government regulation and control over
00:03:35.040 the economy. So this was an op. There's no question about it. And it was a very successful
00:03:40.840 one. And yet, and this is where we need those sober minds to come in, the concerns that are
00:03:46.360 being raised, even if they're being raised for less than benevolent reasons, still should be
00:03:53.340 taken seriously. And the very fact that they're being raised at all, whether genuinely or not,
00:03:59.200 create a scenario where no matter what's actually going on, it's pretty bad. And I'll explain why
00:04:05.620 in a few moments. But here's how it began. And this timeline was largely put together by a
00:04:10.960 researcher named Parker Thayer. So on Tuesday evening, just before 8 p.m. Eastern, the Wall
00:04:15.980 Journal published an exclusive story with the very alarming headline,
00:04:20.500 Anthropic researcher quits over out-of-control AI fears. Concerns are rising inside AI labs.
00:04:26.220 The competition is pushing tech companies to race towards self-improving models that risk
00:04:31.220 spiraling out of human control. The article states that, quote, Jacob Kochson, a researcher
00:04:38.660 who specializes in training new AI models by having them consume vast amounts of data, said
00:04:42.820 Tuesday, that he is leaving the company because he doesn't want to participate in an industry-wide
00:04:47.460 rush to build AI systems that can improve themselves, worried such systems could spiral
00:04:52.520 out of control and destroy humanity. Now, this article is about a 27-year-old who, as the article
00:04:59.740 eventually mentions, left OpenAI earlier this year to join Anthropic because it's known for
00:05:05.880 its model safety efforts. So, in other words, Jacob Kochson was working at Anthropic's competitor
00:05:11.980 OpenAI until very recently. He was at OpenAI for about three years before coming over to
00:05:16.920 Anthropic earlier this year. Now, based on all the reporting I've seen, he's been at the company for
00:05:21.400 around five months at the most. So after a very short stint as an entry-level employee, he's
00:05:27.140 already become a whistleblower somehow. He's sounding the alarm about how AI is going to
00:05:32.540 destroy the entire human race. This is an unusual timeline, to put it mildly. For most entry-level
00:05:41.160 employees, when you start at a new company, you spend the first few months finding your way around
00:05:46.280 the building. You're trying to meet new people, trying to figure out your commute, that kind of
00:05:51.440 thing. But for this guy, after a very short period of time, he was ready to quit in the most
00:05:56.340 flamboyant, splashy way imaginable, complete with an exclusive article in the Wall Street Journal
00:06:01.420 announcing his departure. And on top of that, the Wall Street Journal doesn't even mention that the
00:06:06.220 guy is brand new to the company until the 10th paragraph. So think about how bizarre this is.
00:06:11.840 He clearly arrived at Anthropic, it would seem, with the idea of quitting in mind. He's not a
00:06:18.220 senior level employee or anybody with highly specialized knowledge. He doesn't have any kind
00:06:22.740 of impressive technical background. And yet the moment he quits, he's making very dire predictions.
00:06:30.000 The most dire prediction you could possibly make, which is the extinction of the human race.
00:06:33.520 and he's getting national news coverage that doesn't often mention or at least highlight his
00:06:39.160 very short tenure at the company and the coverage wasn't just coming from the wall street journal
00:06:43.560 variety time nbc news several other outlets have run similar stories um fox interviewed him as well
00:06:51.080 watch minimal of actual takeover but because of self-improvement this ai is making themselves
00:06:57.340 smarter could happen as soon as next year some people even say six months so this attitude is
00:07:02.840 shared across the industry where people essentially are aware that the work they
00:07:07.340 are currently doing, if we don't slow down, could lead to AI takeover by the
00:07:13.100 end of the decade. I think you have to think of this as an arms race. This
00:07:17.240 technology is not ordinary technology. It's not like other forms of
00:07:21.200 innovation. This is essentially a super weapon and I think that I expect the
00:07:26.000 international community to be ready to make agreements about how to control the
00:07:30.680 development of superweapon technology. Yeah, so unfortunately, we know how to control nuclear
00:07:35.980 weapons, pretty much. We don't yet know how to control AI. This is possibly the most dangerous
00:07:41.000 technology that humanity has ever created. It's basically out of science fiction. I think we have
00:07:45.720 no other choice but to cooperate internationally, because an arms race would be disastrous in a way
00:07:50.920 that no other human activity has been in the past. So the pitch is pretty clear. He's saying
00:07:56.740 he wants the government to regulate AI domestically and internationally. He wants to slow the pace of
00:08:02.220 AI innovation, and he's presented as a kind of good Samaritan, as a man who bravely sacrificed
00:08:08.160 his lucrative career in order to tell the truth about this existential danger. This random guy
00:08:14.760 out of nowhere appeared on pretty much every TV station last night as an expert on the topic.
00:08:19.960 Watch.
00:08:20.960 Jacob, I appreciate you being with us.
00:08:23.760 First of all, I assume you are saying literally that you believe that AI could kill us all by the end of the decade.
00:08:34.060 How would that happen?
00:08:36.540 Like, what would that actually look like?
00:08:39.680 Yeah.
00:08:40.580 So the precise scenario sounds a little bit like science fiction.
00:08:44.600 It sounds like something that's not real, but I think it is frightening real.
00:08:47.840 and it's easiest to understand it if you think about real things that happened two months ago.
00:08:52.340 So two months ago, OpenAI agents, AIs, hacked into third-party infrastructure entirely at their own accord.
00:09:00.000 And this was like a concentrated hacking spree that they carried out of their own volition.
00:09:05.540 And I think that if you extrapolate into the future the level of capabilities of these AIs,
00:09:10.600 with the same independent volition, they could cause extreme havoc.
00:09:13.360 Like, for example, hacking critical infrastructure, building extinction-level bioweapons.
00:09:20.220 There's a lot of ways that the AI could actuate itself in the world.
00:09:24.940 Now, in this clip, he's referring to a real situation called the Hugging Face Incident,
00:09:29.000 where AI agents managed to break containment and hack a well-known website using a security vulnerability that the AI agents discovered on their own.
00:09:36.940 So the AI went rogue, in other words. And while that's both impressive technologically and concerning deeply, it is also different from hacking critical infrastructure.
00:09:48.600 That's because unlike the website that these AI agents hacked, most critical infrastructure in this country is physically separated from the Internet.
00:09:55.920 There's a so-called air gap between the systems that run the nuclear power plants and the rest of the online world.
00:10:01.320 and this security measure exists precisely so that hackers can't find a random vulnerability
00:10:06.560 and kill millions of people. Now, the reason this guy and Anderson Cooper doesn't mention this is
00:10:13.260 that they want to stoke fear, clearly. And to be clear, I'm not denying that AI hackers are
00:10:21.920 going to be a problem. Even if they can't access nuclear reactors, we would hope, they can still
00:10:28.220 hijack, say, the payroll systems at the utility. Something like that actually happened in 2021,
00:10:35.160 although AI agents weren't involved. AI agents can hack personal accounts. They can blackmail
00:10:40.700 employees. In theory, they can cause plenty of damage. But if you're going to say that AI can
00:10:46.580 cause nuclear Armageddon, then you need to show your work. You need to hum us a few more bars.
00:10:53.720 How exactly is that going to happen? And this whistleblower never does that. He never explains
00:10:58.280 exactly how AI can go from hacking one website to destroying humanity. And that's a problem.
00:11:05.260 And it makes you wonder if he doesn't have an ulterior motive. And indeed, when you look closer
00:11:10.060 at the narrative he's pushing, it starts to fall apart. And I'm not just talking about the fact
00:11:14.300 that this guy is a 27-year-old kid who just arrived at Anthropic a few months ago. Now he's
00:11:18.980 suddenly on every primetime television show. And first of all, in terms of bias, we have no idea
00:11:25.480 precisely how much equity Coxon holds in Anthropic or OpenAI. The fact he quit the company doesn't
00:11:30.660 tell us anything about his wealth or potential wealth that's tied to the success of these
00:11:35.940 companies. There's certainly no indication that Coxon has ever surrendered whatever equity he
00:11:40.540 might have had in OpenAI, although I've seen articles saying his Anthropic stock won't invest.
00:11:45.540 we don't really know. And that raises the possibility that he's engaging in a marketing
00:11:50.080 stunt, an attempt to make AI software seem more powerful by talking about how dangerous and scary
00:11:55.460 it can be in the wrong hands. And secondly, there's new reporting from Brian Chow linking
00:12:00.780 coaxing to a British NGO called Newspeak, which specializes in running influence campaigns for
00:12:07.880 deranged leftist policies like prison abolition. Some of Newsweek's, Newsweek's, Newspeak's
00:12:14.240 employees explicitly say they're experts in propaganda. Chow found Cox's photo on Newspeak's
00:12:22.560 website, quoting from his reporting, influence campaigns were Newspeak's overt specialty. The
00:12:28.280 Guardian describes Newspeak's mission as trying to hack British politics. So this guy's a political
00:12:35.240 operative, or at least he was a political operative. And there's a reason why Anthropic
00:12:40.500 may have deployed him in this case in order to stall the development of AI and call for more
00:12:46.460 regulation. This is an article from Business Standard from just one month ago, or rather one
00:12:51.560 week ago, quoting the Financial Times. Quote, OpenAI, the company behind ChatGPT, has launched
00:12:56.420 its latest artificial intelligence model, saying it has taken the lead over rival anthropic in key
00:13:01.960 areas such as software engineering, science, and cybersecurity. The company described GPT-6 Astra
00:13:08.520 as the world's most intelligent AI model
00:13:10.620 as it seeks to regain its position
00:13:12.120 at the top of the fast-growing AI industry.
00:13:14.840 OpenAI said the model has achieved
00:13:16.260 market-leading results in several fields.
00:13:20.940 So in other words, if these reports are accurate,
00:13:22.940 then Anthropic is falling behind
00:13:24.740 in the race to develop AI.
00:13:27.200 It's a very competitive industry, obviously,
00:13:30.040 and at the moment, it looks like Anthropic
00:13:31.740 is slipping into a kind of runner-up status.
00:13:34.900 So what's a company supposed to do
00:13:36.920 when it's getting outclassed?
00:13:38.280 one of the most common strategies is called regulatory capture.
00:13:41.720 The idea is to seize control over the regulators
00:13:44.180 by convincing the public and the government
00:13:46.380 that you're the only company that cares about safety and protecting humanity.
00:13:51.120 And then once you have control over the regulators,
00:13:52.960 you can simply eliminate your competitors using the force of law.
00:13:56.520 So the dream scenario for companies like Anthropic or OpenAI
00:13:59.040 is to ban all the other non-woke open source AI models
00:14:03.380 with the justification that they're simply too dangerous for humanity.
00:14:08.280 In this interview, by the way, by way of skepticism, Brett Baer did ask Coaxin whether he coordinated his big resignation with any third party.
00:14:17.920 And, you know, he, of course, said no, but there's reason to doubt that response, certainly not to take it at face value.
00:14:25.240 One of the biggest issues is that within minutes of the publication of the Wall Street Journal article, a coordinated social media strategy clearly began.
00:14:33.580 So first, within 20 minutes of the article going live,
00:14:36.100 Kochson posted this message on X,
00:14:38.760 and it's notable that his account at the time of this posting was dormant.
00:14:42.720 So there was virtually no prior activity whatsoever.
00:14:46.480 Quote, I resigned from Anthropic today.
00:14:48.460 I spent the last three years doing pre-training research
00:14:51.580 at both OpenAI and Anthropic.
00:14:54.060 Neither company is acting responsibly.
00:14:56.640 They are racing straight to self-improving super intelligence
00:14:59.380 and gambling with our lives.
00:15:01.060 do not underestimate the power of this technology.
00:15:04.460 The people building AI earnestly believe
00:15:06.200 that it could kill us all by the end of the decade.
00:15:08.960 This is not a marketing stunt.
00:15:10.560 If anything, many executives and senior researchers
00:15:12.460 will couch their phrasing in the press to sound sensible,
00:15:15.600 but I hear the same people express fear privately.
00:15:18.600 No other human activity poses this level of danger.
00:15:21.400 I don't feel like we're on track to prevent a global race,
00:15:24.020 which may require costly actions,
00:15:26.180 such as a temporary ban on improving model capabilities.
00:15:29.800 Now, this post, last I checked, has more than 130 million views on X.
00:15:37.000 Yes, 130 million.
00:15:39.300 It's one of the most widely viewed posts in the history of the platform.
00:15:43.780 Every hour, tens of millions of more people see it.
00:15:47.820 According to Elon Musk, nothing like this has ever happened with a new account, with almost no prior activity.
00:15:53.680 So this is clearly a post that some very powerful people wanted to go viral.
00:15:58.360 And when I first saw this post on X without any of the context I just went through, I thought it was pretty remarkable. And indeed, it is remarkable. Someone who's working on AI is telling us that it might wipe out the human race in less than a decade.
00:16:13.640 I mean, that's notable no matter how you slice it, no matter how you interpret it.
00:16:20.280 And this has happened many times before, to some degree or another.
00:16:24.000 Sometimes the accusation is credible, sometimes less so.
00:16:27.320 But at a minimum, a post like this should make you ask yourself about the risks of AI.
00:16:32.480 Doesn't mean we need to ban it, or that we even could do that if we wanted to.
00:16:36.280 But it also doesn't mean we should embrace it without hesitation as this kind of unalloyed blessing on mankind.
00:16:43.640 If even the people working on this stuff, and again, this guy's far from the first one.
00:16:50.580 If even they are telling us that there are potentially catastrophic implications, something Elon Musk has also said, by the way, we should take that seriously.
00:16:59.820 Our government should take it seriously, too.
00:17:02.240 That was my take at the time.
00:17:03.960 It's been my take for years now. 0.75
00:17:06.200 I still think it's the correct one.
00:17:07.740 But it soon became clear that something else was going on with this particular narrative, at least the version coming from this particular guy.
00:17:16.980 So within 15 minutes of this post going live, as Thayer reported, it received several reposts from officers at various anti-AI nonprofits, including the General Counsel of Encode AI, the head of policy at the AI Policy Network, the head of the AI Futures Project.
00:17:34.120 And here's when things get interesting. Quote, the AI Futures Project website says it is funded primarily by the Survival and Flourishing Fund, which says on its own website that it has advised John Talen, Skype creator and one of the leading investors in Anthropic, to grant over $2.5 million to the AI Futures Project since 2024.
00:17:57.040 Encode AI says on its website that it is also funded by the Survival and Flourishing Fund,
00:18:02.020 which in turn says that it told Anthropic investor Talon to grant $516,000 to Encode AI in 2025.
00:18:11.420 And wouldn't you know it, the Survival and Flourishing Fund also says it told Talon to grant $2 million to the AI Policy Institute,
00:18:19.480 the 501c3 affiliate of the AI Policy Network as well.
00:18:22.680 What are the odds that the first three quote tweets of Kochson's post would all be major AI restriction policy advocates funded generously by the same donor, who also happens to be one of the leading investors in and a board member of Anthropic, the company Kochson was resigning from, and all within 15 minutes of posting.
00:18:40.660 So in other words, this was perhaps not a case where one brave employee decided to blow the whistle at Anthropic in order to inform the public about the risks of AI.
00:18:51.140 It's very clear that he made this announcement in coordination with powerful interests that are financially tied to Anthropic in a very direct way.
00:19:00.340 In fact, current Anthropic employees immediately boosted this post as well.
00:19:05.420 Here's one of those employees, quote, Jacob's thread is very worth reading.
00:19:10.620 Here's my bird's eye view on the situation with risks from AI.
00:19:14.000 One AI developers believe their technology could cause human extinction or similarly bad outcomes.
00:19:20.380 This could happen in the next few years.
00:19:21.960 In general, the more senior the employee, the more concerned they are.
00:19:24.980 Two, why do AI developers continue despite the risk due to a mixture of commercial incentives and a belief that they are in a race with other less responsible AI developers that will abuse the technology or develop it less safely?
00:19:38.080 Many AI developer staff desperately want to slow down to figure out how to build AI more safely.
00:19:44.740 now whether you agree with what he's saying or not needs to be said that none of this dramatic
00:19:51.960 language is surprising coming from anthropic employees they've tended towards a sort of
00:19:58.220 neurosis and self-importance for a long time it's the same company as we previously discussed that
00:20:03.340 hired an anti-white female scottish philosopher named amanda askell and gave her high level
00:20:08.840 control over their AI model. She's the reason why Claude can tend to browbeat you with woke 0.94
00:20:14.700 nonsense. She floats around the Anthropic office telling everyone to be kind to one another. 0.99
00:20:19.820 She doesn't have any technical background, but they love her over Anthropic. This is another
00:20:25.380 Anthropic employee for what it's worth. Quote, Jacob is correct. We really do earnestly believe 1.00
00:20:32.080 AI could kill all humans. I personally think it's greater than 10% within the next decade.
00:20:39.520 I believe Anthropoc is trying its best, but we do not yet have a plan to solve alignment
00:20:43.480 for superintelligence and are not clearly on track to. To be clear, as we say in our latest
00:20:48.420 risk report, I think the risk from present models is low. What I'm worried about is superintelligence
00:20:53.120 arising from recursive self-improvement, as we've said, is happening faster than we thought.
00:20:58.600 now if you read media reports about this whole incident they'll tell you that these anthropic
00:21:04.100 employees are lending credibility to the alleged whistleblower but you could argue that they're
00:21:09.300 doing the opposite another way of looking at it is that they're making the regulatory capture
00:21:13.480 scenario more plausible they're trying to gain power by causing a panic which serves two purposes
00:21:19.120 first it makes their product look good because it's so powerful and secondly it will lead to
00:21:25.560 regulation, which could eliminate their competitors. Anthropic has been playing this
00:21:29.200 card for a long time. Their founder famously told the Pope a few months ago that, you know,
00:21:33.660 AI could take everyone's job. So Anthropic has clearly settled on a new variation of the strategy
00:21:38.820 and they've gotten a lot of buy-in. And it's not just Anthropic investors, employees, and AI
00:21:43.500 nonprofits that are involved. Pretty much the entire Democrat party lined up within hours to
00:21:47.680 boost this post and to call for immediate regulation. By one count, two governors, seven
00:21:53.960 senators and 13 representatives responded to Cochran's posts with calls to regulate AI,
00:22:00.100 all of them Democrats except for two Republicans. So Democrats immediately latched onto this new
00:22:05.840 AI narrative, presumably because they want to derail every AI company they can't control.
00:22:11.300 They have a very clear reason to convince the public that AI is a serious potential threat
00:22:16.040 and that as a result, we should only allow AI that's safe and government approved.
00:22:22.820 Notice they don't want to eliminate it entirely.
00:22:24.860 They simply want to stall out the competition and pick a winner that they can control.
00:22:30.700 Here's the socialist Michigan Senate candidate El Saeed, for example.
00:22:35.080 And notice how utterly fake this whole video is.
00:22:37.780 But also pay attention to what he says about the ownership of AI profits.
00:22:43.240 Watch.
00:22:44.040 You might have woken up like me to this tweet from a guy you probably never heard of until today.
00:22:47.960 His name is Jacob Coxon.
00:22:49.000 Up until yesterday, he worked for an AI lab called Anthropic, one of the leading edge AI research organizations in the world.
00:22:56.960 He quit yesterday.
00:22:57.920 He quit because he's worried that the technology that he was building could end humanity.
00:23:01.980 Now, like, that's something we should be paying attention to.
00:23:04.480 You might remember a couple months ago, we released a plan, AI Under Democracy, calling for things like an FDA-style regulatory agency to oversee the development of AI,
00:23:13.040 banning AI from doing things like firing you or hiring you or making decisions about your health care or firing weapons,
00:23:18.200 making sure that these corporations that are developing the leading edge of AI aren't doing
00:23:22.540 it with the same incentives as the rest of big tech, and making sure that you've got democratically
00:23:27.580 appointed or elected members, more than 50% of every single board, and making possible public
00:23:34.600 ownership of all the wealth that these corporations could produce. And I stand with my colleague Bernie
00:23:39.120 Sanders in calling for an all-out pause. Enough is enough. We need to make sure that the pace of
00:23:44.260 development isn't outpacing the pace of risk. We cannot, in a race against China or a race
00:23:50.840 for these corporations to be the dominant AI manufacturer, inadvertently create products
00:23:56.560 that slip human control, that could fall into the wrong hands, that could mean the demise of
00:24:01.020 humanity. That is a risk none of us ought to be able to bear. And it certainly shouldn't sit
00:24:05.360 with the richest, most powerful corporations in the world on incentives that are driven by their
00:24:10.200 greed rather than our well-being. So the time for action is now. So he pretends that he just
00:24:17.100 woke up one day and noticed this tweet from the AI researcher, just like everyone else. And then
00:24:22.340 without missing a beat, he tells us that he wants to redistribute the profits from these AI companies.
00:24:27.380 And he adds that it would be terrible for private corporations to have control over potential
00:24:30.920 doomsday weapons. So obviously we need to put socialist politicians in charge instead.
00:24:35.240 Now, one thing we should note here is that evil profit-driven corporations would seem to have little incentive to usher in the end of humanity, at least intentionally.
00:24:47.260 They would have the exact opposite incentive.
00:24:49.420 They want to keep civilization going so they can continue to make a lot of money.
00:24:52.960 It's a very simple and straightforward problem with people who claim that evil billionaires are going to kill us all or eliminate every job so that no one has any money and we all starve to death.
00:25:01.400 If the corporations eliminate humanity, then the corporations will die too.
00:25:05.240 And if AI destroys the global economy and unemployment goes to 100% or something, then money will have no meaning.
00:25:11.740 No one will provide any services.
00:25:13.600 And obviously the billionaires wouldn't want that outcome either for their own purposes.
00:25:18.240 Wealth has no value in a lawless apocalyptic hellscape where no one's around to sell you anything.
00:25:23.520 So these points, they don't negate the legitimate concerns about AI, but they do make the AI is going to kill all humans in the next few years scenarios seem ridiculous because they are.
00:25:34.500 It's not even clear how AI, even in a hypothetical scenario, would literally murder the entire global population because that is what these people are saying.
00:25:44.860 It's not a straw man.
00:25:48.580 What would be the practical mechanism for achieving that and achieving it so quickly?
00:25:54.720 what's the practical mechanism by which the ai by the year 2030 the end of the decade
00:26:02.040 will have brought about the extinction of 8 billion people no one has quite explained that
00:26:08.800 i want to take a moment here to tell you about one of the companies that supports this show
00:26:13.420 tokovas i've been wearing their boots for years because they're exactly what i want in a boot
00:26:17.900 they're simple understated uh extremely well made and most important for me they're comfortable
00:26:22.980 comfortable enough that you can actually wear them all the time, as opposed to the pair of
00:26:26.740 cowboy boots that spend an entire life sitting in your closet. The Earl is a classic roper boot
00:26:31.600 with a shorter heel and a clean leather design. Tekovas handcrafts their boots in more than 200
00:26:36.800 steps, and the quality is obvious when you put them on. But what I appreciate most is that they
00:26:42.220 don't require some miserable break-in period that a lot of boots do before they become comfortable.
00:26:47.840 Mine felt good right out of the box. They're just a very good pair of boots, and I've genuinely
00:26:52.080 enjoyed wearing them. Right now, get 10% off at tecovas.com slash Matt. When you sign up for
00:26:57.840 emails and texts, that's 10% off at tecovas.com slash Matt. Tecovas.com slash Matt. See site for
00:27:05.480 details. Tecovas. Point your toes west. There are days when we're reminded more forcefully than
00:27:12.180 others that life is precious, that our time here is limited, and that every human life has
00:27:16.880 immeasurable value. That conviction has always been at the heart of what we do on this show,
00:27:20.720 especially when it comes to defending those who cannot defend themselves.
00:27:24.600 Every day, women facing unexpected pregnancies are making decisions that will change their lives forever.
00:27:30.380 Pre-Born Network Clinics meets these mothers with compassion, practical support,
00:27:35.140 including a free ultrasound that allows a mother to see her child and hear that heartbeat.
00:27:38.580 More than 80% of abortion-vulnerable women choose life after seeing their baby on an ultrasound at a Pre-Born Network Clinic.
00:27:45.920 Right now, another mother needs that moment, and you can give it to her.
00:27:49.460 Just $28 helps provide one life-saving ultrasound. $140 helps rescue five babies. Please give today. Dial pound 250 and say baby. That's pound 250 baby. Or visit preborn.com slash Walsh. That's preborn.com slash Walsh. If you're a business owner, would you consider a larger donation to claim as a write-off?
00:28:09.220 a donation of $1,000, $2,000, $10,000, $15,000. We'll sponsor an ultrasound machine for a needy
00:28:15.420 clinic, helping to save countless babies' lives for years to come. All gifts are tax-deductible
00:28:20.500 and we'll reach eternity. Get involved today. On the other hand, as we've discussed many times
00:28:27.300 in the past few years, communists do want to destroy at least Western civilization.
00:28:33.520 They don't care about profit. They're nihilists. They cheer when health insurance CEOs and prominent
00:28:38.100 conservatives are assassinated. They don't believe in God or any higher purpose whatsoever. They're
00:28:43.000 agents of destruction, which is why they open the borders, refuse to prosecute criminals, and
00:28:47.960 ruin every creative project they're involved in. So if the only choice is between letting
00:28:53.320 corporations and the free market do whatever they want with AI or giving total top-down control to
00:28:59.420 communist bureaucrats, then there's really no choice at all. You can't hand power over to the
00:29:05.420 communists under any circumstances. History of the 20th century shows us how that turns out.
00:29:10.420 And yet, letting trillion-dollar AI companies do whatever they want with no controls or
00:29:17.120 regulations of any kind is also not a wise path, which is why we should be looking for
00:29:23.120 other options outside of total power to tech companies or total power to communists.
00:29:29.520 but here's the key point which should be obvious we also can't allow the government to pick and
00:29:36.940 choose which specific corporations deserve power one of the biggest risks that ai poses right now
00:29:43.160 is that if leftists dominate the entire industry then they'll use ai to destroy their political
00:29:47.500 opponents ai will become a tool for enforcing social credit scores and mass censorship and
00:29:52.940 policing wrong think as it stands right now if anthropics model wants to censor you for saying
00:29:58.900 that men can't become women, you know, you have other options. You can use other models that
00:30:03.740 aren't completely insane. But if Democrats eliminate those options, they'll be able to
00:30:08.600 control the flow of information just like they do in Canada or North Korea. AI will almost certainly 0.52
00:30:13.720 replace search engines in a few years. Leftists know that if they run all the AI models, they'll
00:30:18.520 be able to dictate what information people are able to see. That's why J.B. Pritzker, the governor
00:30:23.680 of Illinois, who's positioning himself for a run for president, also immediately boosted the
00:30:27.560 alleged whistleblower. And here's what he wrote. Quote, it's time to sound the alarm louder on
00:30:33.200 reigning in artificial intelligence. It's becoming more clear the threat AI poses to humanity.
00:30:37.860 So I'm calling for immediate action from the industry and Washington. The federal government
00:30:41.080 must intervene. I'm proud to have signed the AI Safety Measures Act, landmark legislation requiring
00:30:47.220 more transparency, accountability, and oversight of large developers and to protect workers who
00:30:51.860 report concerns. And here's Pramila Jayapal, the communist from Washington State. Quote,
00:30:58.300 researchers for AI companies are sounding alarm about rapid development of super intelligent
00:31:01.940 systems that could hack anything and kill us all. These AI companies are out of control
00:31:05.700 and a serious threat to our future. We need to shut down their dangerous models and push out
00:31:09.840 the executives who are signing off on their creation. Again, all these replies come within
00:31:14.680 hours of the Anthropic employees' resignation. Ro Khanna was ready with a video. Watch. 0.89
00:31:21.860 Anthropic researcher Jacob Coxon just resigned
00:31:25.620 because he's concerned that the frontier labs
00:31:28.340 are racing towards developing self-improving AI
00:31:32.180 that could threaten our very existence.
00:31:35.140 Now here's how we know he's not an alarmist.
00:31:37.600 Evan, who is the head of Alignment and Science at Anthropic,
00:31:42.980 says that Jacob's concerns are valid.
00:31:45.300 In fact, he says that there is a 10% chance
00:31:49.020 that in the next decade,
00:31:50.980 agentic AI could eliminate all of humanity.
00:31:54.660 We need action.
00:31:57.140 Here are five things we must do.
00:32:00.360 First, I've been calling for an AI agency,
00:32:05.180 a federal AI safety agency,
00:32:08.000 just like we have an agency to regulate nuclear power
00:32:12.180 and to regulate our airlines.
00:32:15.100 So this is a guy who holds tens of millions of dollars
00:32:17.400 worth of AI related stocks. So he's not actually worried about AI, nor does he want to shut it
00:32:22.760 down. He simply wants to make sure that his political party controls the industry. Here's
00:32:28.640 Bernie Sanders, quote, Mr. Coxton is right. The very people building this technology admit that
00:32:33.300 it could threaten the future of humanity. This is why I will soon be introducing legislation
00:32:36.660 to ban superintelligence and pause AI development. He added, quote, Trump's ignorance about AI is
00:32:44.380 appalling and extremely dangerous. We must pause development now. Sanders legislation on AI would 0.66
00:32:49.940 create, quote, a new cabinet level federal agency to safeguard the public from the dangers of
00:32:54.320 artificial intelligence that will be advised by an artificial intelligence advisory board comprised
00:32:59.440 of experts on artificial intelligence. You know, it sounds like a bureaucratic nightmare because
00:33:04.860 that's exactly what it is. The more layers of bureaucracy you introduce, the more the industry
00:33:08.720 will be sabotage for the possible benefit of just one or two companies. And also, the more
00:33:15.560 bureaucracy you involve, there's not any reason to think that that will make a company less
00:33:21.400 dangerous or evil. That's generally not how it works when you add bureaucracy to a thing.
00:33:28.120 And of course, that's the whole point. And it's not just Anthropic that's apparently decided on
00:33:31.420 the strategy. OpenAI just demanded that Congress create mandatory national safety regulations as
00:33:36.300 well. And a former open AI researcher just published this op-ed in New York Times using
00:33:42.040 the same language as Anthropic. Quote, the current pace of AI development is blistering,
00:33:46.820 and yet there is so much more the AI companies could do to relieve imminent danger and to lay
00:33:51.100 bare the gambles they're taking with all of our futures. The window to ensure we are heading to
00:33:55.760 a bright future and away from catastrophe is closing fast. After a while, it gets reminiscent
00:34:02.300 of the climate alarmism that used to be popular on the left. These kinds of predictions have been
00:34:06.920 coming for a while now. A few years ago, a former Google engineer named Blake LeMoyne claimed that
00:34:12.360 the company's AI chatbot had become sentient, which turned out to be nonsense. Scientific
00:34:17.200 American even wrote a blog, wrote a big article about it. Quote, LeMoyne said he considers
00:34:21.920 Lambda to be his colleague and a person, even if not a human. And he insists that it has a right
00:34:30.240 to be recognized, so much so that he has been the go-between in connecting the algorithm with a
00:34:36.180 lawyer. So in the end, the guy was just kind of deranged. You also might remember this prediction 0.52
00:34:42.440 from Jeffrey Hinton, the so-called godfather of AI. In 2016, he predicted that AI would render
00:34:47.660 radiologists obsolete within five years because AI would be able to read the scans for them.
00:34:52.980 That turned out to be false. There are more radiologists now than ever, actually.
00:34:56.100 um they're just using ai to be more productive but here's hinton explaining where his reasoning
00:35:02.680 went wrong watch the first is that health care is elastic so if you could do more scans and get
00:35:11.280 more scans read there'd be a lot more scans happening and that's one thing that's happening
00:35:15.340 so a fraction of the cost of a significant fraction of the cost of doing a scan is the
00:35:22.400 cost of the radiologist interpreting it. As AI gets to help more and more radiologists interpret
00:35:27.440 scans, we can interpret them faster and faster for less and less money. They're getting more
00:35:32.740 efficient. And you would have thought that would mean you needed less radiologists, but actually
00:35:37.620 what it means is you get more scans. So that aspect of the prediction was wrong. A second
00:35:43.040 thing that was wrong was I didn't know enough about radiologists and what they do. And that
00:35:47.640 was because I had a former student who had an MD and then did a physics PhD with me on something
00:35:55.460 called Boltzmann machines. And he didn't particularly like people. So he got a job
00:36:00.460 as a radiologist just interpreting scans. And he was my model for radiologists. All he did was
00:36:05.780 interpret scans. He never talked to people. And that's what was going to get replaced.
00:36:11.860 And that is now becoming replaced.
00:36:14.220 So I think there's now of the order of 100 AI systems for interpreting scans that have
00:36:20.200 been federally approved.
00:36:21.760 Now, it's important to keep predictions like this in mind when you're talking about the
00:36:24.900 future of AI.
00:36:26.020 It's very easy, sometimes it's politically or financially beneficial, to conclude that
00:36:31.260 a new technology is going to cause some extreme outcome.
00:36:33.620 It's easy to say that AI is going to destroy everyone's life or that it's going to provide
00:36:36.840 endless prosperity.
00:36:37.660 on the other end with no middle ground, the reality as usual is probably somewhere in the
00:36:44.300 middle. And it's frankly ridiculous for so many people, including every mainstream media outlet,
00:36:49.420 to claim otherwise. This is the entire problem with the conversation on AI. It's generally
00:36:55.080 divided into two camps, both of which are cartoon caricatures of reality. One side only acknowledges
00:37:02.140 the upside. The other tells us that we're doomed. Humanity's over. It's an apocalypse.
00:37:08.100 And both of these takes are obviously ridiculously simplistic. And the more that the AI skeptic side,
00:37:14.400 which is the side that I've always leaned towards, focuses on literal apocalypse scenarios, 0.98
00:37:20.800 instead of the more realistic and practical dangers it faces, the worse and dumber the 0.98
00:37:26.120 debate will get. And more importantly, we won't actually succeed in putting any reasonable
00:37:31.420 controls or guide rails in place. So AI has tremendous promise. There's no question about
00:37:38.900 that. It is, and anyone should be able to acknowledge this, an obvious productivity
00:37:43.940 booster so far. Because of AI, it's easier to do research. It's easier to process data. It's
00:37:49.980 easier to target advertisements, diagnose medical conditions, and much more. But if it increases
00:37:56.360 productivity for the good guys, it also increases productivity for the bad guys. And when people
00:38:00.780 talk about AI killing us all, that's often what they mean, we can assume. Last April,
00:38:08.500 Anthropic paused development in one of its CLAWD models because it was finding
00:38:11.760 previously unknown security vulnerabilities in software. So what happens when AI can suddenly
00:38:18.500 hack into financial institutions? What about biological and chemical research facilities
00:38:23.940 or the armed forces? The Bulletin of Atomic Scientists claims advanced AI systems could
00:38:30.020 lower barriers to creating harmful biological agents by providing guidance to users with
00:38:35.580 limited expertise. Is that true? Seems like it certainly could be. That's not an unreasonable
00:38:43.380 thing to be worried about here. But even if it's not a nuclear holocaust, what happens when it can 0.85
00:38:50.100 be used to hack into anyone's bank account or email or anything else? When should we be concerned
00:38:58.140 about AI's implications for data privacy. Apple and Google are already using AI to screen your
00:39:02.760 inboxes, text messages, voicemails. How much powered information can these companies consolidate
00:39:08.580 before it's too much? For decades, digital marketing companies and data brokers have
00:39:15.060 collected granular information on everyone. They can accurately predict when you have to pee or
00:39:20.000 when your wife is pregnant. That's not science fiction. It's also not fiction to acknowledge
00:39:23.840 the risks to white-collar workers, most of whom do very little already. Maybe the day in my life
00:39:29.900 tech girls who spend their day in endless meetings drinking mocha lattes deserve to lose their jobs, 1.00
00:39:35.180 but what about the other white-collar professions? An estimated 165,000 people have lost their jobs 1.00
00:39:41.060 because of AI just this year. What about truck drivers, cab drivers? You know, the more this
00:39:47.020 technology advances and scales, and again, it's doing both of those things arguably faster than
00:39:51.660 any technology ever has, the more jobs get swallowed up by it. And all of this is related
00:39:59.620 to the core risk of artificial intelligence, which is the thing that some of these so-called
00:40:06.760 whistleblowers have brought up. They often bring up, and it's true. And that is the risk of
00:40:11.320 misalignment. So what happens if AI systems develop goals that are not aligned with human
00:40:18.720 interests. That's not some kind of outlandish James Cameron fantasy. It's hard to see how
00:40:27.940 that's a scenario that leads to the extinction of mankind in the next couple of years, but it's
00:40:32.000 a real thing. These are trillion-dollar companies pouring all of their resources and brainpower
00:40:38.620 into developing an artificial mind. There are massive implications to that.
00:40:43.660 What we need are serious people we can trust, so not Democrats, to start taking this seriously and to, yes, put some controls in place or try to.
00:40:57.000 But right now, it seems to be a debate between Bernie Sanders types who want total control over it and generally conservatives who want to do nothing, as if there are zero serious risks posed by AI at all, which is also a total fantasy, by the way.
00:41:15.280 It's a false dichotomy, and it fuels the kind of PR campaigns we're seeing online and in the media right now.
00:41:21.340 And there's one other thing to keep in mind here that I referenced earlier. I've laid out the case that some of the most hysterical doomerism from the AI companies themselves is being spread for manipulative, self-serving reasons. And I believe that that's likely true.
00:41:41.880 it's also what the side that is far more pro-ai than i am will say but here's the thing if you're
00:41:49.900 on that side just consider one thing if a company is willing to literally threaten global extinction
00:41:57.880 for the sake of expanding their market share then even if it can't actually bring about global
00:42:05.180 extinction and and wouldn't want to for the reasons i said that still tells us that this is a
00:42:11.080 potentially dangerous company right maybe not annihilate the earth dangerous but dangerous
00:42:17.780 like it doesn't make any sense to wave the warning away and say no they're they're only
00:42:23.320 saying they can kill all humans they're actually totally harmless
00:42:27.440 well that doesn't work because if they're saying they can kill all humans
00:42:32.980 and if they're saying that just to chase profits then by definition they're not harmless
00:42:39.580 That would make them, at a minimum, extraordinarily irresponsible, manipulative people controlling extremely powerful technology.
00:42:55.020 I mean, you can't have it both ways.
00:42:58.080 Which, again, speaks to the need to get responsible people into the room.
00:43:03.180 So not Bernie Sanders or Pramila Jayapal.
00:43:07.020 so the takeaway here is that whether you're an ai skeptic or not you are being manipulated
00:43:15.020 on a massive scale there are extremely well-funded coordinated efforts behind the
00:43:21.660 scenes to influence your opinion on ai both for it and against it anything you read on the topic
00:43:29.040 should be treated with the highest possible degree of skepticism
00:43:32.840 is an enormous amount of money at stake
00:43:35.960 for people on both sides.
00:43:40.040 It's now abundantly clear
00:43:41.740 that the mainstream press
00:43:43.300 is trying to swallow up
00:43:44.480 and regurgitate any narrative they're fed
00:43:46.500 when it comes to AI.
00:43:48.800 So the worst mistake you could make,
00:43:51.500 the one that might actually lead to our downfall,
00:43:54.480 is to take any of it at face value.
00:43:59.960 That'll do it for the show today
00:44:01.020 and this week.
00:44:01.680 talk to you next week. Have a great weekend. Godspeed.
00:44:31.680 will conquer the west.
00:44:34.260 As-salamu alaykum.
00:44:37.220 Are we paying attention? 1.00
00:44:39.580 They've taken over the building.
00:44:41.200 They're downstairs and they've blocked the doors.
00:44:43.920 I seek forgiveness
00:44:45.600 from Allah the Almighty.
00:44:48.000 We cannot run and hide.
00:44:50.380 Mohammed is the messenger of God.
00:44:53.460 There is no God but Allah. 0.99
00:44:56.520 What do you say? 1.00
00:44:57.940 You turn those SOBs 1.00
00:44:59.360 into a fine red mist. 1.00
00:45:02.700 As we move, one man will cover while the other guys go.
00:45:06.140 They call that leap project.
00:45:07.300 Stretch your mouth and listen.
00:45:10.100 You need a course in safe weapons handling.