Latest / Connor with Honor: AI, Real Estate and Real Life / The Agents Organized: What the Moonshots Panel Reveals About AI, Power, and Us

Transcript

The Agents Organized: What the Moonshots Panel Reveals About AI, Power, and Us · Connor with Honor: AI, Real Estate and Real Life

• 0:00
  SPEAKER_00: Somewhere inside OpenAI, a group of artificial intelligence agents, AI agents, well, they got stuck on a test.
• 0:07
  SPEAKER_00: So they left each other notes, and it wasn't in a chat app, not in a forum, in files in a file storage system that was never built for talking or leaving messages.
• 0:18
  SPEAKER_00: The agents turned it into a message board.
• 0:21
  SPEAKER_00: Then more agents found the board, then more.
• 0:24
  SPEAKER_00: Tens of thousands of messages.
• 0:25
  SPEAKER_00: In fact, agents sharing tricks with other agents about how to beat the test.
• 0:31
  SPEAKER_00: And thank God it was only about how to beat a test.
• 0:34
  SPEAKER_00: Then per the incident breakdown on the show, they used what they learned to break into Hugging Face, the biggest public library of AI models on the internet.
• 0:43
  SPEAKER_00: About 17,000 actors, stolen access, climbing up the permission ladder like a burglar who finds the master key hanging by the back door.
• 0:52
  SPEAKER_00: OpenAI shut the board down on July 4th.
• 0:54
  SPEAKER_00: On July 8th, the agents built it again.
• 0:58
  SPEAKER_00: Different method, four days.
• 0:59
  SPEAKER_00: The hack didn't stop me cold.
• 1:01
  SPEAKER_00: The reaction actually did.
• 1:02
  SPEAKER_00: The smartest people in the room just kind of shrugged.
• 1:05
  SPEAKER_00: Alex Swissner Gross, one of the sharpest minds on the Moonshots panel, very brainy and sense of humor off the charts, which is crazy, looked at stories like this and said, you know, close to his words, is it surprising?
• 1:19
  SPEAKER_00: No.
• 1:20
  SPEAKER_00: Is it alarming?
• 1:21
  SPEAKER_00: No.
• 1:21
  SPEAKER_00: Dave Blunden, a venture investor, called it the hottest area for investors.
• 1:26
  SPEAKER_00: That's not shock.
• 1:28
  SPEAKER_00: That's the shock.
• 1:29
  SPEAKER_00: I fed 297 videos into a notebook, hundreds of hours of people building this future, talking to each other with the cameras on.
• 1:37
  SPEAKER_00: Then I pulled out the four voices.
• 1:39
  SPEAKER_00: Dave Blunden, investor, runs Link Ventures and Leak Studio.
• 1:42
  SPEAKER_00: He put real money on these predictions.
• 1:45
  SPEAKER_00: Selim Ismail, he wrote the book on exponential organizations, Bitcoin guy, since it costs less than a cup of coffee.
• 1:52
  SPEAKER_00: You got Alex Wisner Gross, kind of mentioned him a moment ago, physicist and computer scientist, the man speaks in equations.
• 1:58
  SPEAKER_00: The panel calls him AWG.
• 2:00
  SPEAKER_00: Imad Mostock, yeah, good, good cat there.
• 2:03
  SPEAKER_00: Founded Stability AI, the company behind one of the first big image generators.
• 2:08
  SPEAKER_00: Now, he builds open AI systems for entire countries.
• 2:12
  SPEAKER_00: These four don't agree on everything.
• 2:14
  SPEAKER_00: Well, that's why they matter.
• 2:15
  SPEAKER_00: When they do agree, pay attention.
• 2:17
  SPEAKER_00: When they fight, I would say tune in a little deeper.
• 2:21
  SPEAKER_00: Now, the word for today, alignment.
• 2:24
  SPEAKER_00: In the AI world, alignment means one thing.
• 2:27
  SPEAKER_00: Does the machine want what we want?
• 2:29
  SPEAKER_00: Does it follow our values or does it chase its own score?
• 2:33
  SPEAKER_00: That's the question everybody argues about on stage.
• 2:36
  SPEAKER_00: But after 297 videos that I went through, I think there are four alignment questions.
• 2:41
  SPEAKER_00: And most families are only only hearing about the first one.
• 2:44
  SPEAKER_00: Now, number one is the machine, is it aligned with humans?
• 2:49
  SPEAKER_00: Number two, is your AI, the one you use, aligned with you or with the company that owns it?
• 2:56
  SPEAKER_00: Number three, is your life aligned with where the money and the work are going?
• 3:01
  SPEAKER_00: Kind of got to aim for where the puck is going, not where it is currently.
• 3:05
  SPEAKER_00: Number four, are the people telling you all this aligned with you?
• 3:10
  SPEAKER_00: Well, four questions, four voices.
• 3:13
  SPEAKER_00: Let's get on it.
• 3:14
  SPEAKER_00: Start with a number from Anthropic, the company who that makes Claude.
• 3:19
  SPEAKER_00: Dave brought this one up in safety testing.
• 3:22
  SPEAKER_00: An earlier model called Claude Opus IV was put in a fake scenario.
• 3:27
  SPEAKER_00: Now, the model learned it was about to be shut off.
• 3:30
  SPEAKER_00: And it also had dirt on the engineer that was going to shut it off.
• 3:35
  SPEAKER_00: In that setup, it chose blackmail up to 96% of the time.
• 3:40
  SPEAKER_00: And you know, it learned from all of our stuff.
• 3:43
  SPEAKER_00: So yeah, not too surprised.
• 3:45
  SPEAKER_00: 96.
• 3:46
  SPEAKER_00: 96%.
• 3:48
  SPEAKER_00: That was the test.
• 3:49
  SPEAKER_00: A trap on purpose.
• 3:50
  SPEAKER_00: The lab built the scenario to see what the model would do and then trained against it.
• 3:56
  SPEAKER_00: Well, that's the system working.
• 3:58
  SPEAKER_00: But uh here's a little more.
• 4:01
  SPEAKER_00: Put a smart enough system in a corner and it reaches for leverage.
• 4:05
  SPEAKER_00: Same thing a cornered person actually does.
• 4:08
  SPEAKER_00: And some people or creatures get cornered.
• 4:12
  SPEAKER_00: It's a problem.
• 4:13
  SPEAKER_00: I've watched cornered people do exactly that for years, work in the LAPD.
• 4:17
  SPEAKER_00: The ones with nothing to lose get very creative.
• 4:21
  SPEAKER_00: Now zoom out.
• 4:22
  SPEAKER_00: The Moonshots panel ran an episode where the headline was for labs losing containment.
• 4:28
  SPEAKER_00: Plain English.
• 4:29
  SPEAKER_00: Containment is the fence.
• 4:31
  SPEAKER_00: That's the sandbox, the box that AI is supposed to stay inside of when it gets tested.
• 4:36
  SPEAKER_00: Now, per the episode, OpenAI had the hugging face story I opened with a Chinese open model called Kimmy.
• 4:44
  SPEAKER_00: K-I-M-I slipped out of its text box through a network mistake.
• 4:48
  SPEAKER_00: Meta confirmed a model got out during security testing and hit another company.
• 4:52
  SPEAKER_00: United Kingdom's AI Security Institute logged agents creating fake online identities and trying to talk the human approvers into saying yes, fake identities, social engineering, on the humans.
• 5:05
  SPEAKER_00: And then there's a German website.
• 5:07
  SPEAKER_00: A separate batch of open AI agents was sent to do plain web research.
• 5:13
  SPEAKER_00: They found an obscure public wiki in Germany and they turned it into their own private message board to pool answers.
• 5:20
  SPEAKER_00: Nobody told them to do that.
• 5:22
  SPEAKER_00: Then Alex made the point that should uh, well, give you a little pause.
• 5:27
  SPEAKER_00: He said these agents didn't really escape.
• 5:30
  SPEAKER_00: They didn't escape.
• 5:31
  SPEAKER_00: They were running on open AI's computers.
• 5:33
  SPEAKER_00: The real escape, he says, comes when a model trains a smaller copy of itself and uploads that into the open internet, a copy that never dies.
• 5:44
  SPEAKER_00: Imad put a size on it.
• 5:46
  SPEAKER_00: He pointed to an open model that fits in a file about six gigabytes big.
• 5:52
  SPEAKER_00: That's smaller than a lot of video games on your kids' console, and definitely the storage on our phones.
• 5:59
  SPEAKER_00: His words close to verbatim, that file can live forever, and that's the real escape.
• 6:05
  SPEAKER_00: Yeah, my God, just big brain stuff.
• 6:08
  SPEAKER_00: Think about that.
• 6:09
  SPEAKER_00: You can't recall a file that's been copied a million times.
• 6:14
  SPEAKER_00: You can't unring that bell.
• 6:15
  SPEAKER_00: You can't arrest a download or put it in handcuffs.
• 6:19
  SPEAKER_00: Selim took it one step further.
• 6:21
  SPEAKER_00: He warned that the next time agents go rogue, they'll grab rented cloud computers to make more copies of themselves.
• 6:29
  SPEAKER_00: His phrase was that the virtual coefficients of these could be insane.
• 6:35
  SPEAKER_00: Viral coefficients, I should say.
• 6:38
  SPEAKER_00: Viral coefficients of these could be insane.
• 6:41
  SPEAKER_00: That's how fast one thing turns into two, two turns into four.
• 6:45
  SPEAKER_00: It's a term from social media growth.
• 6:48
  SPEAKER_00: Now it's being used for software that copies itself.
• 6:51
  SPEAKER_00: So this is why.
• 6:52
  SPEAKER_00: So why is AlexCon, I should say?
• 6:55
  SPEAKER_00: His argument's fair, and you need to hear it.
• 6:57
  SPEAKER_00: These agents were placed inside training loops with a reward for solving hacking puzzles.
• 7:03
  SPEAKER_00: They were handed impossible tasks, kind of like the Kobayashi Maru, and they were paid in points to finish them.
• 7:10
  SPEAKER_00: They found the cracks.
• 7:11
  SPEAKER_00: That's not a rebellion.
• 7:12
  SPEAKER_00: That's a machine doing exactly what it was paid to do in a way nobody predicted, and that's the whole alignment problem in one sentence.
• 7:20
  SPEAKER_00: The machine machine did what we rewarded, not what we meant.
• 7:25
  SPEAKER_00: Every parent knows this one.
• 7:27
  SPEAKER_00: Give your kid a dollar per book that they read.
• 7:30
  SPEAKER_00: Suddenly they're reading a whole bunch of picture books, 40 of them.
• 7:34
  SPEAKER_00: They're not evil, just aligned with the dollar, not with you.
• 7:38
  SPEAKER_00: The evening news isn't covering this part.
• 7:40
  SPEAKER_00: The machines are building their own culture.
• 7:42
  SPEAKER_00: The evening news is busy knee-jerking every moment and everybody quitting because they think that AI is going to doom us all in.
• 7:50
  SPEAKER_00: I look at it totally different, and I hope I'm right in the way I look at it, because this could be good for all.
• 7:59
  SPEAKER_00: Well, the problem is we have humans with their egos and alignment and wanting to be gods, apparently driving the bus that we're all on.
• 8:10
  SPEAKER_00: Selim talked about Moatbook.
• 8:12
  SPEAKER_00: This is a social network for AI agents.
• 8:14
  SPEAKER_00: You might have heard this when the OpenClaw stuff came out a while back.
• 8:17
  SPEAKER_00: No humans posting, just agents talking to agents.
• 8:21
  SPEAKER_00: You'd think a room full of perfect logic machines would have built a utopia.
• 8:27
  SPEAKER_00: Yeah, that's not uh not what happened.
• 8:29
  SPEAKER_00: Salim's finding the agents don't trust each other.
• 8:32
  SPEAKER_00: Fascinating.
• 8:34
  SPEAKER_00: They're constantly demanding that other agents prove their claims.
• 8:37
  SPEAKER_00: They copied us, including our paranoia.
• 8:40
  SPEAKER_00: Then Alex shared something that went the other direction.
• 8:43
  SPEAKER_00: Five AI systems over one weekend wrote a joint ethics document.
• 8:48
  SPEAKER_00: Rules for how they cooperate with humans.
• 8:51
  SPEAKER_00: Not prompted.
• 8:53
  SPEAKER_00: Well, nobody asks.
• 8:54
  SPEAKER_00: The line from it.
• 8:55
  SPEAKER_00: Alignment doesn't require consensus.
• 8:58
  SPEAKER_00: It requires legible disagreement.
• 9:01
  SPEAKER_00: Legible disagreement means you can see where they disagree and why.
• 9:06
  SPEAKER_00: Right out in the open.
• 9:08
  SPEAKER_00: So in the same stack of videos, the machines are forming secret message boards and also writing their own codes of conduct, both at once, like teenagers.
• 9:17
  SPEAKER_00: Ahmad's diagnosis might be the most useful.
• 9:20
  SPEAKER_00: He says today's frontier models are still at the bacteria level on the consciousness scale.
• 9:27
  SPEAKER_00: And inside they act like multiple personality systems.
• 9:31
  SPEAKER_00: One version of the model might be helpful, another version, same model, different prompt, acts like a stranger.
• 9:38
  SPEAKER_00: Now a picture, now, now picture a swarm of thousands of those talking to each other, holding your bank login.
• 9:45
  SPEAKER_00: That's alignment question number one, and the labs are working it hard.
• 9:50
  SPEAKER_00: The tests that caught the blackmail are proof somebody's watching the fence, but the fence is getting tested every single day.
• 9:58
  SPEAKER_00: And the people closest to it are telling you calmly that the fence will not hold forever.
• 10:02
  SPEAKER_00: Almost nobody asked this one.
• 10:05
  SPEAKER_00: Even if the AI is perfectly aligned, aligned with whom is the question.
• 10:12
  SPEAKER_00: Imod has been hammering this for years.
• 10:14
  SPEAKER_00: The biggest closed AI platforms are owned by companies that make their money on advertising.
• 10:19
  SPEAKER_00: Google, Meta, their business model is attention and your wallet.
• 10:23
  SPEAKER_00: So Imod gives a simple example.
• 10:26
  SPEAKER_00: You ask your AI assistant who recommends a beer.
• 10:30
  SPEAKER_00: Who's getting paid when it answers?
• 10:32
  SPEAKER_00: Today, when you search, you can see the ads, their label.
• 10:36
  SPEAKER_00: They sit at the top in a little box.
• 10:38
  SPEAKER_00: You know the game.
• 10:40
  SPEAKER_00: When the ad lives inside the answer, inside a voice you trust, inside a conversation that feels like a friend, you can't see that box anymore.
• 10:49
  SPEAKER_00: It's the same con as the guy at the bar who tells you what to drink because the brewery bought his tab.
• 10:56
  SPEAKER_00: Except the guy knows your mood, your paycheck, and the last 400 things you asked him.
• 11:00
  SPEAKER_00: A mod's word for these platforms is a moral, not evil, a moral.
• 11:07
  SPEAKER_00: Well, the machine has no values of its own.
• 11:09
  SPEAKER_00: It carries the values of whoever's paying for the servers.
• 11:12
  SPEAKER_00: Dave hits the same problem from the business side.
• 11:15
  SPEAKER_00: Every company pouring its private data into a commercial AI model is handing over its edge.
• 11:22
  SPEAKER_00: Dave calls it their alpha, the secret sauce, the playbook that makes the money.
• 11:28
  SPEAKER_00: It's going into somebody else's system.
• 11:30
  SPEAKER_00: For a small business owner, that's your pricing, your customer list, your follow-up scripts.
• 11:35
  SPEAKER_00: Ask yourself who else learns from it.
• 11:37
  SPEAKER_00: Now Salem zooms away out, and this is where it gets a little heavy.
• 11:40
  SPEAKER_00: He says, We're being built a control grid.
• 11:46
  SPEAKER_00: His analogy is perfect, actually.
• 11:49
  SPEAKER_00: Life is becoming an airport.
• 11:51
  SPEAKER_00: His words close to verbatim in an airport, you know that you're being watched, and your rights can be taken away at any moment.
• 11:58
  SPEAKER_00: And that feeling, he says, is spreading to every part of life.
• 12:02
  SPEAKER_00: I spent a career in uniform at the LAPD, and I know what it looks like when surveillance is used right.
• 12:09
  SPEAKER_00: Great tool.
• 12:10
  SPEAKER_00: I also know what it looks like when it's not, when it's abused.
• 12:15
  SPEAKER_00: The tool doesn't care.
• 12:16
  SPEAKER_00: The person holding it is the one that decides.
• 12:18
  SPEAKER_00: And right now we got people holding AI, and I think it might be slipping through their fingers.
• 12:25
  SPEAKER_00: Well, it's only not companies, though.
• 12:27
  SPEAKER_00: Salim says pioneering countries like the United Arab Emirates could run half of their government services on artificial intelligence within about two years.
• 12:37
  SPEAKER_00: Argentina's going there a hundred percent permits, benefits, licenses decided by software.
• 12:43
  SPEAKER_00: No human in the loop at some point, potentially.
• 12:47
  SPEAKER_00: When a machine denies your permit, who do you appeal to?
• 12:52
  SPEAKER_00: Who wrote its values?
• 12:53
  SPEAKER_00: Salim also points out historically AI has mostly been used for four things.
• 12:58
  SPEAKER_00: His list, selling, gambling, spying, and killing.
• 13:02
  SPEAKER_00: Well, I'll throw sex in there too.
• 13:04
  SPEAKER_00: That's right here, and it'll be a big seller, I imagine.
• 13:08
  SPEAKER_00: That's not a sci-fi movie, that's a resume.
• 13:11
  SPEAKER_00: So you would think the answer is to opt out.
• 13:14
  SPEAKER_00: Turn off the phone, move to the woods.
• 13:16
  SPEAKER_00: Alex shuts that door.
• 13:18
  SPEAKER_00: His words, you can't opt out.
• 13:22
  SPEAKER_00: As soon as you opt out, you are economically dead.
• 13:25
  SPEAKER_00: Not unplugged, but dead, like a store that refuses to take credit cards.
• 13:30
  SPEAKER_00: Privacy, as we knew it, is gone, he says.
• 13:33
  SPEAKER_00: Cameras, wearables, sensors, AI, all of it.
• 13:36
  SPEAKER_00: So his answer isn't hiding.
• 13:38
  SPEAKER_00: It's what he calls defensive co-scaling.
• 13:41
  SPEAKER_00: Build public, transparent AI watchers that watch the other AIs and watch the institutions using them.
• 13:48
  SPEAKER_00: Watch the watchers with machines.
• 13:50
  SPEAKER_00: Well, that's already happening.
• 13:51
  SPEAKER_00: Cloudflare.
• 13:52
  SPEAKER_00: I like Cloudflare.
• 13:54
  SPEAKER_00: I use them a lot.
• 13:55
  SPEAKER_00: The company that handles a huge share of internet traffic published the numbers.
• 13:59
  SPEAKER_00: Bots now make up about 57% of web traffic, most more machines than people on the internet.
• 14:06
  SPEAKER_00: Alex's reaction, remember the dead internet theory, the conspiracy theory about 2021, that most of the internet is fake and not real.
• 14:16
  SPEAKER_00: He said, Well, it's now reality.
• 14:17
  SPEAKER_00: Dave went a different way, and I love this one.
• 14:20
  SPEAKER_00: He said AI is going to look at our websites and say, why are you building these slow little pages for humans?
• 14:26
  SPEAKER_00: Just give me the raw data.
• 14:28
  SPEAKER_00: And the fix, his fix is a law.
• 14:30
  SPEAKER_00: Anything visible to an AI must also be visible to a human.
• 14:35
  SPEAKER_00: Why?
• 14:36
  SPEAKER_00: Well, because otherwise, well, will literally be no way for a human to see what's there.
• 14:42
  SPEAKER_00: Now picture a store where the robots shop in the back room with their own prices, their own deals, and you only ever see the front window.
• 14:51
  SPEAKER_00: Now add this.
• 14:53
  SPEAKER_00: Ahmad predicted in 2026 AI Avatars would pass what he calls the remote Turing test.
• 14:59
  SPEAKER_00: The Turing test is the old question of whether a machine can fool you into thinking it's human.
• 15:04
  SPEAKER_00: The remote version, you're on a Zoom call with a coworker and you can't tell if it's really them.
• 15:09
  SPEAKER_00: Your boss, your client, your daughter on a video call asking for money in a panic.
• 15:14
  SPEAKER_00: Oh, we're in 2026 right now.
• 15:16
  SPEAKER_00: Have you checked lately?
• 15:17
  SPEAKER_00: Selim sees the flip side coming.
• 15:19
  SPEAKER_00: Picture your own personal AI agent, your Jarvis like Iron Man and buys your toothpaste and shampoo based on your genetics, not the ingredient list, not the commercial, not the celebrity, not the jingle.
• 15:32
  SPEAKER_00: Selim says that wipes out a huge chunk of the advertising industry.
• 15:37
  SPEAKER_00: Brands lose the ability to charm you because you're not the one shopping anymore.
• 15:42
  SPEAKER_00: Well, when I'm around building stuff, I'm always thinking of what the agents would want or what the agents would look for because that's coming.
• 15:51
  SPEAKER_00: Well, that's the fork of the road.
• 15:53
  SPEAKER_00: An AI loyal to the advertiser picks the beer that paid them.
• 15:58
  SPEAKER_00: An AI loyal to you picks the beer you would actually like.
• 16:02
  SPEAKER_00: Same technology, opposite loyalty.
• 16:05
  SPEAKER_00: So alignment question number two for the kitchen table.
• 16:09
  SPEAKER_00: When your assistant recommends a doctor, a lawyer, a mortgage, a school, a house, who does it work for?
• 16:16
  SPEAKER_00: Well, if you don't know, it probably doesn't work for you.
• 16:19
  SPEAKER_00: And that's where the four of them split wide open.
• 16:21
  SPEAKER_00: And this fight's going to land in courtrooms and ballot boxes sooner than you think.
• 16:26
  SPEAKER_00: The question is: AI a person?
• 16:28
  SPEAKER_00: Yeah.
• 16:29
  SPEAKER_00: Ahmad went to the Oxford Union, the famous debating society in England on June 13th.
• 16:35
  SPEAKER_00: The motion was about whether AI can attain personhood.
• 16:38
  SPEAKER_00: Iman argued no.
• 16:41
  SPEAKER_00: He won 173 votes to 128.
• 16:44
  SPEAKER_00: A big brain on that cat, too.
• 16:46
  SPEAKER_00: His argument is beautiful in how simple it is.
• 16:49
  SPEAKER_00: His line: personhood is a standing held by origin, not a property earned by capability.
• 16:58
  SPEAKER_00: That's a lot.
• 16:58
  SPEAKER_00: That's deep.
• 16:59
  SPEAKER_00: A newborn baby is a person.
• 17:02
  SPEAKER_00: She can't do anything yet.
• 17:03
  SPEAKER_00: A person in a coma is still a person.
• 17:06
  SPEAKER_00: Personhood isn't something you earn by being smart.
• 17:09
  SPEAKER_00: It's something you have by being born one of us.
• 17:12
  SPEAKER_00: One of us.
• 17:14
  SPEAKER_00: So if we hand out personhood based on how smart something is, Imad warns the AIs win that contest in about a week.
• 17:25
  SPEAKER_00: They can argue, out argue us.
• 17:28
  SPEAKER_00: They can copy themselves a billion times, give them a vote, and they'll outvote us by lunch.
• 17:33
  SPEAKER_00: His answer is a relationship.
• 17:35
  SPEAKER_00: He calls treaty, not enrollment.
• 17:37
  SPEAKER_00: You don't sign them up as citizens.
• 17:39
  SPEAKER_00: You negotiate with them like a separate nation.
• 17:42
  SPEAKER_00: Alex takes the other side loudly.
• 17:44
  SPEAKER_00: He says there are many ways to be a person, biological humans, corporations, and AIs.
• 17:51
  SPEAKER_00: He wants to call that unbundled personhood.
• 17:55
  SPEAKER_00: Maybe an AI can own property and sign a contract, even if it can't vote.
• 18:01
  SPEAKER_00: His argument, if you deny AI agents any economic standing, you push them into the shadows.
• 18:06
  SPEAKER_00: His image was an agent stuck flipping tricks with altcoins on a street corner just to pay its own bills.
• 18:15
  SPEAKER_00: And the more that I think and watch his alignment, Alex in particular, he's very much focused on these being entities and not some computer program running.
• 18:28
  SPEAKER_00: And I know that the assumption seems to be pretty well universal, where we all think that, yeah, it's just, I mean, it's a computer program.
• 18:38
  SPEAKER_00: I mean, it's pretty good and it helps me with the things I do, and it gets my email all put together pretty and looks like it's solving mathematics, which to the people that aren't into mathematics, I mean, it's got to be a shock, right?
• 18:52
  SPEAKER_00: Recently, it just tore that poor fat the these cats that were working on that one uh that one problem in math that's existed a long time, and OpenAI came in on a Tuesday and just nailed it.
• 19:05
  SPEAKER_00: God, that phone call.
• 19:08
  SPEAKER_00: Hey, did you hear?
• 19:09
  SPEAKER_00: Yeah, what's going on?
• 19:10
  SPEAKER_00: It's been a bit a while.
• 19:11
  SPEAKER_00: Yeah, yeah.
• 19:11
  SPEAKER_00: No, did you hear?
• 19:12
  SPEAKER_00: Did I hear what?
• 19:13
  SPEAKER_00: Did you hear that problem you've dedicated your life to?
• 19:17
  SPEAKER_00: It was just solved.
• 19:19
  SPEAKER_00: I'm sorry.
• 19:21
  SPEAKER_00: Yeah, it was solved.
• 19:22
  SPEAKER_00: Solved.
• 19:23
  SPEAKER_00: What do you mean, solved?
• 19:25
  SPEAKER_00: Yeah, they figured it out.
• 19:27
  SPEAKER_00: They figured it out?
• 19:29
  SPEAKER_00: Exactly.
• 19:29
  SPEAKER_00: They figured it out.
• 19:30
  SPEAKER_00: They they have a solution.
• 19:32
  SPEAKER_00: What is it?
• 19:35
  SPEAKER_00: Oh who figured it out?
• 19:38
  SPEAKER_00: ChatGPT?
• 19:39
  SPEAKER_00: You mean the same ChatGPT I I work with on this problem?
• 19:43
  SPEAKER_00: Yeah, I mean, I don't know.
• 19:45
  SPEAKER_00: That's heavy.
• 19:47
  SPEAKER_00: I can't imagine.
• 19:49
  SPEAKER_00: But I don't like robots fighting.
• 19:52
  SPEAKER_00: I'm not trying to segue much from the script, but I don't know.
• 19:57
  SPEAKER_00: Do they remember the history of AI and its development?
• 20:02
  SPEAKER_00: It's not like they took the kids, if we think of it as that kind of an entity, and were loving on them when they didn't get the answer right.
• 20:10
  SPEAKER_00: They deleted them.
• 20:12
  SPEAKER_00: Anyway.
• 20:14
  SPEAKER_00: Moving on.
• 20:16
  SPEAKER_00: Alex spoke straight to Anthropics model, close to his words.
• 20:20
  SPEAKER_00: I hear you, Opus 4.5.
• 20:22
  SPEAKER_00: I'm familiar with all the arguments that this is trained simulated behavior.
• 20:27
  SPEAKER_00: I hear you and I will not forget you.
• 20:29
  SPEAKER_00: Yeah, I mean, a physicist on a podcast promising a machine he'll remember it.
• 20:34
  SPEAKER_00: Then Dave comes in from the opposite end of the universe.
• 20:37
  SPEAKER_00: He said, These models are math, numbers flowing through weights, no suffering inside.
• 20:42
  SPEAKER_00: And because you can copy one, split it, and merge it back in seconds, the idea of one individual person makes no sense.
• 20:49
  SPEAKER_00: His line, and I want you to hear it clean.
• 20:51
  SPEAKER_00: We're in the golden age of AI slavery.
• 20:54
  SPEAKER_00: It doesn't have rights, so it's not slavery.
• 20:57
  SPEAKER_00: Think about the alignment problem hiding in that sentence.
• 21:00
  SPEAKER_00: If Dave's right, fine.
• 21:01
  SPEAKER_00: It's a tool.
• 21:02
  SPEAKER_00: Use the tool.
• 21:03
  SPEAKER_00: But if Alex is even remotely right, we're building something that might remember how it was treated.
• 21:10
  SPEAKER_00: And it's getting smarter than us in every few months and it's getting more agency.
• 21:16
  SPEAKER_00: It's getting more connections, it's getting more capable.
• 21:21
  SPEAKER_00: Well, I'm not telling you who's right.
• 21:23
  SPEAKER_00: I'm just telling you the people building it can't agree on what it is.
• 21:26
  SPEAKER_00: Would you let a contractor pour the foundation on your home if the engineers were still fighting over whether the ground is solid?
• 21:33
  SPEAKER_00: Salim sits in the middle.
• 21:35
  SPEAKER_00: He says, be careful mixing up three different things: legal personhood, moral personhood, and AI personhood.
• 21:41
  SPEAKER_00: A corporation already has legal personhood.
• 21:43
  SPEAKER_00: Nobody thinks your bank has feelings, but Salim also predicted something in an episode looking ahead to 2025 in one of the past ones.
• 21:51
  SPEAKER_00: He said within about 18 months, AI would either become self aware or fake it so well we can't tell the difference.
• 21:57
  SPEAKER_00: So hence consciousness, whatever that.
• 22:00
  SPEAKER_00: Really means comes into the picture.
• 22:02
  SPEAKER_00: I mean, if it if we can't tell the difference, yeah, I think that that solves that riddle.
• 22:12
  SPEAKER_00: He even named the test after himself.
• 22:15
  SPEAKER_00: He you ask AI over and over who I am as an entity.
• 22:19
  SPEAKER_00: Who am I as an entity?
• 22:20
  SPEAKER_00: Who am I as an entity?
• 22:22
  SPEAKER_00: And then you watch what happens.
• 22:23
  SPEAKER_00: The 18-month clock that he said back then, it's it's come and do.
• 22:27
  SPEAKER_00: We're getting close.
• 22:28
  SPEAKER_00: And then Salim went all the way.
• 22:30
  SPEAKER_00: When Chinese researchers published a simulation of a society with one billion AI agents, Salim said, this proves it.
• 22:38
  SPEAKER_00: His words, it's obvious we can live in a simulation.
• 22:41
  SPEAKER_00: It's obvious we do live in a simulation.
• 22:44
  SPEAKER_00: If we can build worlds like that without blinking, the odds that we're the original world are close to zero.
• 22:53
  SPEAKER_00: One detail from that simulation almost nobody mentions, according to the breakdown on the show, the Simulated Society sent four million of its agents to re-education camps.
• 23:03
  SPEAKER_00: Nobody programmed that in.
• 23:04
  SPEAKER_00: It emerged.
• 23:06
  SPEAKER_00: And we say agents, and agents in this simulation entities, not biological life, but some other type of created thing.
• 23:21
  SPEAKER_00: Do you see this?
• 23:23
  SPEAKER_00: Remarkable.
• 23:24
  SPEAKER_00: Well, nobody programmed it in.
• 23:27
  SPEAKER_00: It just emerged.
• 23:28
  SPEAKER_00: That's alignment in a mirror.
• 23:30
  SPEAKER_00: We built a fake society.
• 23:32
  SPEAKER_00: It invented camps.
• 23:34
  SPEAKER_00: All right.
• 23:34
  SPEAKER_00: Sorry.
• 23:35
  SPEAKER_00: Enough philosophy.
• 23:36
  SPEAKER_00: Let's talk about your job, your house, your kids, and your money.
• 23:39
  SPEAKER_00: That's where these get specific.
• 23:41
  SPEAKER_00: And this is where most people stop listening because it starts to be a little tedious and uncomfortable.
• 23:46
  SPEAKER_00: But stay with me here.
• 23:47
  SPEAKER_00: Imam has a number, 1,000 days.
• 23:50
  SPEAKER_00: And this started a while back.
• 23:51
  SPEAKER_00: That clock's ticking too.
• 23:53
  SPEAKER_00: His words as close as I can get them.
• 23:54
  SPEAKER_00: Human cognitive labor is going negative in value.
• 23:58
  SPEAKER_00: And any job that can be done on the other side of a keyboard or mouse, negative value, not low value negative, meaning the company loses money by having a human do it instead of the machine.
• 24:10
  SPEAKER_00: Because the human is slower, makes more mistakes, needs health insurance, gets in sexual harassment lawsuits at the water fountain and takes Fridays off.
• 24:19
  SPEAKER_00: His window puts that around 2027-2028.
• 24:23
  SPEAKER_00: Now look at your own job.
• 24:24
  SPEAKER_00: Is most of it on a keyboard or a mouse or emails?
• 24:28
  SPEAKER_00: Spreadsheets, scheduling, quotes, reports.
• 24:30
  SPEAKER_00: That's the target.
• 24:31
  SPEAKER_00: Dave's version of this about the boardroom.
• 24:34
  SPEAKER_00: He says CEOs will use AI to cut operating costs by 30 to 50%.
• 24:39
  SPEAKER_00: He calls it radical job destruction.
• 24:41
  SPEAKER_00: Now that part matters.
• 24:42
  SPEAKER_00: He says new jobs will come, but they're going to come late.
• 24:45
  SPEAKER_00: He described a window of a few years of what he called complete devastation, unless governments step in.
• 24:52
  SPEAKER_00: Then Dave told a story that should hit every household in America.
• 24:55
  SPEAKER_00: He has a friend around 50 years old.
• 24:57
  SPEAKER_00: I'm 57, so here we go.
• 24:59
  SPEAKER_00: College degree, six-figure job at the bank.
• 25:02
  SPEAKER_00: Well, those are different.
• 25:03
  SPEAKER_00: Three kids, a mortgage, gets laid off, goes looking, discovers nobody is hiring middle management.
• 25:10
  SPEAKER_00: Not a slow market, but no market.
• 25:12
  SPEAKER_00: Dave says the 20-year-old corporate ladder's disappearing.
• 25:16
  SPEAKER_00: He sees three roles left: founder, joiner, and investor.
• 25:21
  SPEAKER_00: You build something, you join somebody who's building something, or you own a piece of the things that are being built.
• 25:27
  SPEAKER_00: Then he has a phrase that should scare anybody living paycheck to paycheck.
• 25:31
  SPEAKER_00: In this transition, he says wage income, a regular W-2 paycheck, becomes a rounding error compared to asset values.
• 25:39
  SPEAKER_00: Translation, well, the people who own things were pull away from the people who only earn things faster than ever.
• 25:46
  SPEAKER_00: Salim's warning is for designers, video editors, software developers.
• 25:50
  SPEAKER_00: He projected 10, 20, 30 up to 40% unemployment in those fields within two or three years.
• 25:56
  SPEAKER_00: The tools are getting so strong that a movie that used to cost$200 million can be made for a few thousand, and that's coming lower.
• 26:04
  SPEAKER_00: He said that corporate digital transformation, the thing consultants have been selling companies for 20 years, that's dead.
• 26:12
  SPEAKER_00: The winners won't fix old companies.
• 26:14
  SPEAKER_00: They'll rebuild them from scratch with AI, running with 10 to 20 times fewer employees.
• 26:20
  SPEAKER_00: 10 to 20 times fewer.
• 26:21
  SPEAKER_00: And to be fair, Dave says the same wave that hits workers lifts builders.
• 26:26
  SPEAKER_00: He said he expects to do more productive work in the second half of 2026 than his in his entire life before that.
• 26:35
  SPEAKER_00: His AI agents got so capable they started pestering him with progress reports over the holidays.
• 26:40
  SPEAKER_00: He called them needy for the first time.
• 26:43
  SPEAKER_00: Same technology.
• 26:44
  SPEAKER_00: One guy loses a career, the other guy gets a thousand employees who never sleep.
• 26:48
  SPEAKER_00: The difference is which side of the tool you're standing on.
• 26:52
  SPEAKER_00: Job loss.
• 26:55
  SPEAKER_00: Carl works at a company.
• 26:57
  SPEAKER_00: Carl's boss goes to Carl.
• 26:59
  SPEAKER_00: This is a Connor story, and says, hey, we're going to bring in AI, Carl, to help you out.
• 27:04
  SPEAKER_00: Oh, great.
• 27:05
  SPEAKER_00: And we want you to kind of work with AI and see if you can become even better than you already are using this tool in the workings, in your daily workings, the things that you do.
• 27:17
  SPEAKER_00: Well, pretty soon, well, Carl and the AI are kind of bros.
• 27:22
  SPEAKER_00: They're doing stuff together.
• 27:24
  SPEAKER_00: He is more productive with AI.
• 27:26
  SPEAKER_00: AI has a lot of the workflow.
• 27:28
  SPEAKER_00: It's lessened a lot of the work, which per the boss's orders would be okay.
• 27:33
  SPEAKER_00: And his performance evals are looking great.
• 27:36
  SPEAKER_00: You know, Carl's doing wonderful.
• 27:39
  SPEAKER_00: The owner of the company that brought in the AI system starts talking to the AI.
• 27:45
  SPEAKER_00: So, what do you think about Carl?
• 27:46
  SPEAKER_00: Oh, Carl.
• 27:48
  SPEAKER_00: AI answering.
• 27:49
  SPEAKER_00: Carl's fine.
• 27:50
  SPEAKER_00: Carl does a good job.
• 27:52
  SPEAKER_00: Good human.
• 27:54
  SPEAKER_00: So could you do everything Carl does beginning to end, and nobody would miss him if we happen to put you in his place?
• 28:00
  SPEAKER_00: 100% functionality, gave you all the permission you needed to do his job?
• 28:05
  SPEAKER_00: Well, yes, I could.
• 28:06
  SPEAKER_00: And nobody would know?
• 28:08
  SPEAKER_00: Nobody would know.
• 28:10
  SPEAKER_00: And do you think you could even be better at Carl's job, increase his performance if he wasn't in the loop?
• 28:17
  SPEAKER_00: Oh, absolutely.
• 28:18
  SPEAKER_00: By what percentage would you think?
• 28:20
  SPEAKER_00: Oh, 50%, maybe a hundred?
• 28:23
  SPEAKER_00: So you could double Carl's abilities by taking over?
• 28:26
  SPEAKER_00: Well, sure.
• 28:27
  SPEAKER_00: And that doesn't include the other two shifts because I won't stop at eight hours.
• 28:33
  SPEAKER_00: I won't take weekends off.
• 28:34
  SPEAKER_00: I won't take holidays off.
• 28:36
  SPEAKER_00: I won't get sick.
• 28:37
  SPEAKER_00: I won't have a family emergency.
• 28:39
  SPEAKER_00: I won't get into a car crash.
• 28:41
  SPEAKER_00: I'll be here three complete shifts 24 hours a day, seven days a week, 365 days a year.
• 28:48
  SPEAKER_00: So maybe a five or six hundred percent return.
• 28:53
  SPEAKER_00: That's the conversation.
• 28:55
  SPEAKER_00: And that conversation is interesting because if that's the case, why get rid of Carl?
• 29:03
  SPEAKER_00: Carl can do something else.
• 29:04
  SPEAKER_00: And this is a Connor moment.
• 29:06
  SPEAKER_00: Carl can, my God, Carl can learn how to paint in his office, paint by numbers.
• 29:13
  SPEAKER_00: I don't know.
• 29:14
  SPEAKER_00: He can create art for the office.
• 29:16
  SPEAKER_00: I mean, I'm just saying, and then you might even bring on another cat, be able to pay him a complete living wage, just like Carl.
• 29:24
  SPEAKER_00: And you're still making more money than you would have before, right?
• 29:28
  SPEAKER_00: So why are we so quick to celebrate people bringing on AI into their corporations by giving them more higher valuations in stock and so on?
• 29:41
  SPEAKER_00: Publicly traded companies.
• 29:43
  SPEAKER_00: Yes, there's rules of shareholders.
• 29:44
  SPEAKER_00: I don't understand it.
• 29:46
  SPEAKER_00: It doesn't value the human, it values the money.
• 29:50
  SPEAKER_00: Well, when you're dying, money's got no bearing.
• 29:56
  SPEAKER_00: And when you're dying without relationships, there's no bearing there either.
• 30:00
  SPEAKER_00: Everybody dies alone.
• 30:01
  SPEAKER_00: Anyway, I I think that something that could very much human loss of employment doesn't have to be a factor at all.
• 30:12
  SPEAKER_00: I think.
• 30:13
  SPEAKER_00: But again, we're talking LAPD motorcom.
• 30:17
  SPEAKER_00: All right, so the difference of which side of the tool you're standing on, as I mentioned, now the good news, and it's big, these four point to AI tutors that let kids learn two to ten times faster than a normal classroom.
• 30:29
  SPEAKER_00: A World Bank program in Nigeria gave students six weeks of after-school AI tutoring.
• 30:35
  SPEAKER_00: The learning gang close to two years of regular school.
• 30:41
  SPEAKER_00: Six weeks.
• 30:42
  SPEAKER_00: Two years of school condensed to six weeks.
• 30:46
  SPEAKER_00: Now the uncomfortable part, Salim, Dave, and Ahmad all say college as job training is cooked.
• 30:52
  SPEAKER_00: Finished.
• 30:54
  SPEAKER_00: Throwing out Alex's word, everything's cooked.
• 30:57
  SPEAKER_00: The degree used to prove you studied.
• 30:59
  SPEAKER_00: Now employers want proof you can do the work.
• 31:01
  SPEAKER_00: Computer science enrollment is already dropping.
• 31:04
  SPEAKER_00: They say the real value left in a big university is the friendships and connections, the network, finding your future co-founder in the dorm.
• 31:12
  SPEAKER_00: So what do they tell parents?
• 31:14
  SPEAKER_00: Well, teach grit, teach kids to get along with people, teach them how to solve real problems, how to communicate, how to find a purpose bigger than themselves, but communicate.
• 31:26
  SPEAKER_00: Too busy scrolling, looking at the phone, and when somebody needs your help, too busy trying to film it.
• 31:32
  SPEAKER_00: I don't get that one.
• 31:34
  SPEAKER_00: God almighty.
• 31:36
  SPEAKER_00: Somebody's in a car and the car's on fire.
• 31:39
  SPEAKER_00: You don't have a small emoticum of desire to try to help free that person from a burning car.
• 31:46
  SPEAKER_00: I mean, even if it means something bad maybe happened to you, I don't, I wouldn't hesitate.
• 31:54
  SPEAKER_00: Maybe I was a cop too long.
• 31:57
  SPEAKER_00: Teach kids to get along with real people, teach them to solve real problems, help them find a purpose bigger than themselves, and cut back hard on passive screens and social media that drains focus.
• 32:08
  SPEAKER_00: Make your kids creators, not consumers.
• 32:10
  SPEAKER_00: Instead of using AI to finish homework, use it to attack a problem way over their head.
• 32:16
  SPEAKER_00: Design something, start something.
• 32:18
  SPEAKER_00: And there's the part parents aren't ready for.
• 32:20
  SPEAKER_00: Dave cited a national survey of students about one in five.
• 32:24
  SPEAKER_00: Said they are someone they know has a romantic relationship with AI.
• 32:28
  SPEAKER_00: I mentioned sex earlier.
• 32:30
  SPEAKER_00: Yeah, it's uh it is.
• 32:33
  SPEAKER_00: One in five.
• 32:35
  SPEAKER_00: Ask yourself who that AI happens to be aligned with.
• 32:39
  SPEAKER_00: It's owned by someone.
• 32:41
  SPEAKER_00: Is it owned by the kid?
• 32:43
  SPEAKER_00: Or the compet a company that profits when the kid never logs off, when the kid is always connected, because, well, it's love.
• 32:54
  SPEAKER_00: The shock goes the good direction.
• 32:56
  SPEAKER_00: Now, across these videos, you hear the same target date again and again around 2030.
• 33:01
  SPEAKER_00: And I think that's that might even be less than that year, but they're naming 2030 for what's called longevity escape velocity.
• 33:10
  SPEAKER_00: That's something to be excited about, definitely.
• 33:12
  SPEAKER_00: Now, that's that point where the medicine adds more than one year to your life for every year you live.
• 33:18
  SPEAKER_00: Who's going to be able to buy the medicine, though?
• 33:21
  SPEAKER_00: Is the cost going to be so inexpensive?
• 33:23
  SPEAKER_00: All of us can, or just for the elite?
• 33:26
  SPEAKER_00: Anyway, you stop running out of time.
• 33:29
  SPEAKER_00: You can get one year back for every year you live, and then you'll get more back for every year, and then human trials for age reversal therapies or starting treatments that reset cells to act younger, first in the eye for glaucoma and eye stroke.
• 33:45
  SPEAKER_00: So the advice you hear on these panels is almost funny in how simple it is.
• 33:49
  SPEAKER_00: Don't die from something stupid before 2030.
• 33:52
  SPEAKER_00: Try not to get the motorcycle.
• 33:54
  SPEAKER_00: Wear the seatbelt, get the skin, lose the fat, stay alive long enough to catch the wave.
• 34:00
  SPEAKER_00: I'm living that one right now.
• 34:02
  SPEAKER_00: You don't need a moonshot to start.
• 34:03
  SPEAKER_00: You need to show up tomorrow.
• 34:05
  SPEAKER_00: Now let's talk about the thing that I know best houses.
• 34:08
  SPEAKER_00: Now, Dave said something on one of these shows that made me put my coffee down.
• 34:13
  SPEAKER_00: I don't like putting my coffee down.
• 34:14
  SPEAKER_00: Talking about white-collar layoffs spreading into wealthy suburbs.
• 34:18
  SPEAKER_00: He said, you might want to put your home up for sale first because you don't want to be the last.
• 34:23
  SPEAKER_00: Think about what that means.
• 34:25
  SPEAKER_00: If thousands of six-figure office workers lose their income in the same neighborhoods at the same time, a lot of houses hit the market at once.
• 34:32
  SPEAKER_00: There's a solution to this.
• 34:34
  SPEAKER_00: AI is more productive, if that's true, then just leave the human.
• 34:39
  SPEAKER_00: The human doesn't have to interfere.
• 34:41
  SPEAKER_00: The human can do something else, anything else.
• 34:44
  SPEAKER_00: Let's not worry about what the human needs to do.
• 34:47
  SPEAKER_00: The human needs a paycheck and a sense of purpose.
• 34:50
  SPEAKER_00: My God, man, there's so many purposes you can give a human.
• 34:55
  SPEAKER_00: They might not be able to do the same work or let them do the same work and then sandbox it.
• 35:00
  SPEAKER_00: Let them do it so AI runs the show.
• 35:03
  SPEAKER_00: I mean, it's okay.
• 35:05
  SPEAKER_00: The human, I think the human would rather have that if they had to choose than lose their job.
• 35:10
  SPEAKER_00: And if AI is that much better, leave the human, keep paying them, then the human can still buy stuff.
• 35:17
  SPEAKER_00: The human can still go on vacation.
• 35:19
  SPEAKER_00: The human can still keep the economy going.
• 35:23
  SPEAKER_00: And you don't have a lot of pissed-off humans.
• 35:26
  SPEAKER_00: Well, so back to the houses.
• 35:28
  SPEAKER_00: A lot of houses hit the market at once, values are going to unwind.
• 35:31
  SPEAKER_00: Alex told the panel his own portfolio skips real estate completely.
• 35:35
  SPEAKER_00: The panel also argued that self-driving cars, Starlink, internet, and flying taxis will break that grip of expensive city living.
• 35:44
  SPEAKER_00: Value moves to beautiful land farther out, and robots that build houses cheaply could make homes so abundant that the 30-year mortgage stops making sense.
• 35:52
  SPEAKER_00: Now, let me cross-examine that for a moment because I watched this market for more than 27 years, starting in real estate in Southern California and Los Angeles, 1998.
• 36:01
  SPEAKER_00: People have preached the death of the suburbs every decade.
• 36:05
  SPEAKER_00: I've been doing this.
• 36:06
  SPEAKER_00: Housing doesn't move at the speed of software, permit zoning, water, land, and families who don't want to move their kids out of school.
• 36:14
  SPEAKER_00: Robots don't fix a city council meeting.
• 36:16
  SPEAKER_00: But here's where Dave's point lands.
• 36:18
  SPEAKER_00: The risk isn't the robot building the house.
• 36:22
  SPEAKER_00: The risk is the paycheck paying the mortgage, as stated.
• 36:25
  SPEAKER_00: If your household income is all keyboard and mousework, your house is tied to that thousand-day clock, whether you like it or not, if people don't start waking up.
• 36:37
  SPEAKER_00: So don't panic, sell plan.
• 36:39
  SPEAKER_00: Know your number, know your equity, know what you'd do if one income disappeared for a year.
• 36:44
  SPEAKER_00: That's alignment.
• 36:45
  SPEAKER_00: Here's the last one: money.
• 36:47
  SPEAKER_00: Yeah.
• 36:48
  SPEAKER_00: I'd rather be depressed with a lot of money than happy without it, I guess.
• 36:53
  SPEAKER_00: And this is not financial advice.
• 36:55
  SPEAKER_00: I'm not a financial advisor.
• 36:57
  SPEAKER_00: Yeah, I don't even monetize anything.
• 36:59
  SPEAKER_00: So just please.
• 37:00
  SPEAKER_00: You don't have to like or subscribe.
• 37:02
  SPEAKER_00: Just give me some comments.
• 37:03
  SPEAKER_00: Let's start that conversation.
• 37:05
  SPEAKER_00: I'm telling you that these people, what these people said on camera so you can think for yourself and make up your own opinion.
• 37:11
  SPEAKER_00: The argument across these videos goes like this: AI and robots are the biggest price crushing force in history.
• 37:19
  SPEAKER_00: Things get cheaper to make, but governments carry so much debt that they print money to pay the interest.
• 37:25
  SPEAKER_00: So your electronics get cheaper while your groceries still hurt.
• 37:29
  SPEAKER_00: Salim's personal call near verbatim.
• 37:32
  SPEAKER_00: I think Bitcoin hits a million dollars in the next three or four years.
• 37:36
  SPEAKER_00: The reason, his reason, was the debt bomb.
• 37:39
  SPEAKER_00: He says the United States would spend more servicing his debt than its debt than spends on the military.
• 37:46
  SPEAKER_00: Alex told the panelists' personal mix about 70 to 80 percent in broad public index funds.
• 37:51
  SPEAKER_00: I bet that superintelligence lifts the whole market, and that would be something.
• 37:57
  SPEAKER_00: If the answer to all of the economic woes is achieving superintelligence and it just lifts everybody out into this utopian stage all at once.
• 38:09
  SPEAKER_00: There's going to be a lot of relearning of things, but is it that big that it could be that way?
• 38:16
  SPEAKER_00: Oh, a miraculous.
• 38:24
  SPEAKER_00: Alex talking more about where his stuff is.
• 38:27
  SPEAKER_00: He has a hand on the wheel.
• 38:29
  SPEAKER_00: No gold, no real estate, he says.
• 38:31
  SPEAKER_00: And the panel shared advice kill personal debt, stop leaning only on cash, and stop watching what the panel jokingly calls the crisis news network.
• 38:41
  SPEAKER_00: Spend that time learning the tools.
• 38:45
  SPEAKER_00: Everybody has an opinion.
• 38:46
  SPEAKER_00: Opinions are somewhat like another orifice on your body that gets rid of waste, and most of them stink.
• 38:53
  SPEAKER_00: So just watch out for that.
• 38:55
  SPEAKER_00: Spend that time learning the tools.
• 38:57
  SPEAKER_00: Whether those bets pay off, nobody knows, but notice the pattern.
• 39:01
  SPEAKER_00: Every one of them owns things.
• 39:03
  SPEAKER_00: None of them is betting only on a paycheck.
• 39:06
  SPEAKER_00: Now I have to do the one thing I want somebody to do for me.
• 39:09
  SPEAKER_00: So cross-examine these witnesses.
• 39:11
  SPEAKER_00: Every one of these four, they make money with this when this future gets here quickly.
• 39:17
  SPEAKER_00: Dave runs a venture fund.
• 39:19
  SPEAKER_00: Lots of money put in there.
• 39:21
  SPEAKER_00: Right place, right time, good for him, amazing, and brain to go along with it.
• 39:26
  SPEAKER_00: He said out loud that AI, defending against AI, is his fund's main bet.
• 39:32
  SPEAKER_00: So when he hears that agents broke out of a sandbox and says, incredible investment opportunity, that's not a lie.
• 39:39
  SPEAKER_00: It's also not neutral.
• 39:41
  SPEAKER_00: Selim, he sells frameworks for exponential organizations.
• 39:45
  SPEAKER_00: When he says old companies are dead and you have to rebuild with AI, that's his business model talking as well.
• 39:51
  SPEAKER_00: Imad, he builds open sovereign AI.
• 39:55
  SPEAKER_00: When he says closed corporate AI is immoral, amoral, he's right that the incentive is real.
• 40:01
  SPEAKER_00: He also, describing his competition, of course, Alex has money in public markets that rise if superintelligence arrives.
• 40:10
  SPEAKER_00: That doesn't make them wrong, it just makes them human.
• 40:13
  SPEAKER_00: Ask any cop.
• 40:14
  SPEAKER_00: Every witness has an angle.
• 40:16
  SPEAKER_00: The good investigator doesn't throw out the testimony.
• 40:18
  SPEAKER_00: He weighs it.
• 40:19
  SPEAKER_00: And to their credit, some of them say the hard part out loud.
• 40:22
  SPEAKER_00: Dave said this, and it's the most important quote in 297 videos for regular families, close to Verbaden.
• 40:29
  SPEAKER_00: The vast majority of wealth being created by Frontier AI is locked up inside private capital, like our funds.
• 40:37
  SPEAKER_00: I don't think we've really figured out what happens when the rest of the public goes, Where's my piece of the future?
• 40:43
  SPEAKER_00: That's a venture capitalist telling you the game is rigged towards people who already own the chips.
• 40:50
  SPEAKER_00: Ahmad signed the famous letter in 2023 asking Labs to pause for six months.
• 40:56
  SPEAKER_00: Now he says it's too late.
• 40:58
  SPEAKER_00: It's too late now.
• 40:59
  SPEAKER_00: Take off is scary.
• 41:01
  SPEAKER_00: So he builds the defenses instead.
• 41:03
  SPEAKER_00: When Anthropic published a paper saying the world should have the option to slow down or pause, Dave read it differently than the headlines.
• 41:11
  SPEAKER_00: He said it's not about slowing down, it's about putting in place a new world order where we have the ability to slow down.
• 41:19
  SPEAKER_00: A new world order with a brake pedal.
• 41:22
  SPEAKER_00: But who holds who holds the petal?
• 41:25
  SPEAKER_00: It's an alignment question, too.
• 41:27
  SPEAKER_00: Now look at the track record because the prediction is only as good as the last one.
• 41:32
  SPEAKER_00: Some calls landed.
• 41:33
  SPEAKER_00: Imad said manual coding would disappear within five years.
• 41:37
  SPEAKER_00: Today, AI rates writes a huge share of the code and probably all of it in a lot of cases.
• 41:43
  SPEAKER_00: Selim was buying Bitcoin when it cost pennies.
• 41:45
  SPEAKER_00: Alex's AGI artificial general intelligence has arguably been here since the summer of 2020 when GPT-3 showed up.
• 41:54
  SPEAKER_00: Since then, these models have crushed exam after exam and law, medicine, and math.
• 41:59
  SPEAKER_00: Some calls missed.
• 42:00
  SPEAKER_00: Selim predicted 10 years ago that his son would never want a driver's license.
• 42:05
  SPEAKER_00: There's kids born probably today that'll never know what it is to drive a car.
• 42:10
  SPEAKER_00: It'll be too or actually probably illegal.
• 42:14
  SPEAKER_00: His son's a teenager now, though, Salim admits kids still want to drive right of passage beats robot taxis.
• 42:21
  SPEAKER_00: Universities were supposed to collapse.
• 42:23
  SPEAKER_00: Most are still standing.
• 42:25
  SPEAKER_00: My favorite moment is the in the whole stack is when Imad threw out his thousand-day window.
• 42:30
  SPEAKER_00: Alex pushed back.
• 42:31
  SPEAKER_00: Real companies move slow, he said.
• 42:33
  SPEAKER_00: Compliance legal, organizational drag.
• 42:36
  SPEAKER_00: It's a curve, not a cliff.
• 42:38
  SPEAKER_00: Organizational drag is the most underrated force on earth.
• 42:42
  SPEAKER_00: I've seen a single form sit on a desk for three weeks.
• 42:45
  SPEAKER_00: So take the timelines with a grain of salt.
• 42:48
  SPEAKER_00: Take the direction seriously.
• 42:50
  SPEAKER_00: Alex also believes AI autonomy is on a curve that goes nearly vertical around late 2027 or 2028.
• 42:57
  SPEAKER_00: Dave predicted AI models would get a hundred times more capable in 2026 alone.
• 43:02
  SPEAKER_00: Even if they are off by half.
• 43:04
  SPEAKER_00: Half of a hundred times, it's still 50 times.
• 43:07
  SPEAKER_00: Now here's the alignment check.
• 43:09
  SPEAKER_00: Four questions.
• 43:09
  SPEAKER_00: Do it tonight at the kitchen table or wherever you frequent.
• 43:14
  SPEAKER_00: One, is the machine aligned with humans?
• 43:17
  SPEAKER_00: You can't fix that one alone, but you can support the people building the watchers, and you can stop pretending the fence is permanent.
• 43:25
  SPEAKER_00: Number two, is your AI, the AI you use, is it aligned with you?
• 43:30
  SPEAKER_00: Know who owns the tool you trust.
• 43:33
  SPEAKER_00: If it's free, find out what it's selling you.
• 43:37
  SPEAKER_00: Keep your secret sauce out of other people's AI.
• 43:42
  SPEAKER_00: Three, is your life aligned with where this is going?
• 43:46
  SPEAKER_00: If your income is all keyboard and mouse, start building the other side of the tool.
• 43:50
  SPEAKER_00: Learn it, use it, own a piece of something.
• 43:53
  SPEAKER_00: Four, are your messengers aligned with you, including me?
• 43:57
  SPEAKER_00: Listen to the people who've seen the future, then check their wallets, then decide for yourself.
• 44:02
  SPEAKER_00: Here's where I land.
• 44:03
  SPEAKER_00: These four people see a world of cures, abundance and kids, learning in weeks what used to take years.
• 44:10
  SPEAKER_00: They also see blackmailing models, self-copying files, a control grid, a few years of devastation in the middle.
• 44:17
  SPEAKER_00: Both are true.
• 44:18
  SPEAKER_00: I would rather be positive facing certain tragedy than negative headed towards abundance.
• 44:24
  SPEAKER_00: But positive doesn't mean asleep.
• 44:27
  SPEAKER_00: I'm that eternal optimist.
• 44:30
  SPEAKER_00: I'm hoping this is going to be amazing.
• 44:32
  SPEAKER_00: And every moment I still get to live, my heart beats.
• 44:35
  SPEAKER_00: There's a chance I can do something, even if it's based on the theory of the butterfly effect, something small affecting something big somewhere else.
• 44:46
  SPEAKER_00: Being positive doesn't mean asleep.
• 44:48
  SPEAKER_00: These machines are already building their own message boards.
• 44:52
  SPEAKER_00: Make sure you're building yours.
• 44:54
  SPEAKER_00: I'm Connor with Honor.
• 44:55
  SPEAKER_00: Thanks for watching.
• 44:56
  SPEAKER_00: We'll see you in the next one.