Latest / Connor with Honor: AI, Real Estate and Real Life / AI when it comes to political alignment and what the doomers said they took back

Transcript

AI when it comes to political alignment and what the doomers said they took back · Connor with Honor: AI, Real Estate and Real Life

• 0:00
  SPEAKER_00: Ninety-nine United States Senators agreed on something last summer.
• 0:03
  SPEAKER_00: Ninety-nine.
• 0:04
  SPEAKER_00: Now think about how insane that sentence is.
• 0:07
  SPEAKER_00: We live in a country where those hundred people couldn't agree on lunch.
• 0:11
  SPEAKER_00: They can't agree on what day it is, and ninety-nine of them stood up and voted the same way on artificial intelligence.
• 0:18
  SPEAKER_00: Now here's the part that should make you sit up a bit.
• 0:21
  SPEAKER_00: It was not Republicans on one side and Democrats on the other, it was Ed Markey, the progressive from Massachusetts, standing shoulder to shoulder with Marcia Blackburn, the conservative from Tennessee.
• 0:32
  SPEAKER_00: Those two agree on nothing.
• 0:35
  SPEAKER_00: Nothing.
• 0:36
  SPEAKER_00: And they wrote an amendment together to kill a 10-year ban on states regulating artificial intelligence.
• 0:42
  SPEAKER_00: It passed 99-1.
• 0:44
  SPEAKER_00: The only man who voted no was a Republican from North Carolina.
• 0:47
  SPEAKER_00: So today we're going to talk about the strangest political story in America right now.
• 0:52
  SPEAKER_00: The parties have swapped sides on artificial intelligence.
• 0:56
  SPEAKER_00: The people who normally scream about regulation are demanding it.
• 0:59
  SPEAKER_00: The people who normally scream about corporate powers are defending it.
• 1:03
  SPEAKER_00: Bernie Sanders and Steve Bannon are using the exact same word to describe the exact same men.
• 1:10
  SPEAKER_00: And there's$125 million sitting on a table trying to make sure you never notice.
• 1:16
  SPEAKER_00: I'm going to give you both sides today, the real ones, not the cartoon versions.
• 1:21
  SPEAKER_00: I use this technology every single day, all day long, 17-hour days, it's made me faster, and I have never been 43 years of writing code.
• 1:31
  SPEAKER_00: Never been this good.
• 1:32
  SPEAKER_00: And I'm not here to tell you it's the devil either, but I'm here to not tell you or sell you a dream while somebody quietly rewires their own future.
• 1:41
  SPEAKER_00: So let's go.
• 1:43
  SPEAKER_00: Start with this right.
• 1:45
  SPEAKER_00: Now, this is where it's going to get weird.
• 1:48
  SPEAKER_00: Steve Bannon, on his own show called Artificial Intelligence, likely the most dangerous technology in the history of mankind.
• 1:55
  SPEAKER_00: Not concerning, not worth watching.
• 1:58
  SPEAKER_00: The most dangerous technology in human history.
• 2:01
  SPEAKER_00: And when he talks about the men building it, he does not sound like a conservative talking about business.
• 2:06
  SPEAKER_00: He sounds like a labor organizer.
• 2:09
  SPEAKER_00: He said, and I want you to hear his exact words, and I quote, I am a capitalist.
• 2:14
  SPEAKER_00: This is not capitalism.
• 2:16
  SPEAKER_00: This is corporatism and crony capitalism.
• 2:20
  SPEAKER_00: Majority Taylor Green came out against a president's own executive order, making it easier to build data centers, a Republican against a Republican president, on a business permitting rule.
• 2:32
  SPEAKER_00: Her reasoning was that competing with China does not mean becoming China by trampling states' rights and burning through the water supply.
• 2:39
  SPEAKER_00: Nancy Mace, Republican, South Carolina, called for a statewide moratorium on data center construction, a moratorium on building things, coming out of the mouth of a Republican.
• 2:51
  SPEAKER_00: And Tucker Carlson sat across from Sam Altman and told him to his face that the unsettling part of his technology is that he does not know what it stands for.
• 3:02
  SPEAKER_00: But it clearly stands for something.
• 3:04
  SPEAKER_00: Altman said that there's nothing divine about it.
• 3:06
  SPEAKER_00: Tucker did not look convinced.
• 3:08
  SPEAKER_00: Now flip that board.
• 3:09
  SPEAKER_00: Bernie Sanders released a report last October warning that AI and automation could eliminate close to 100 million American jobs in a decade.
• 3:18
  SPEAKER_00: His answer was a 32-hour work week with no cut in pay, a tax on companies that replace the worker with a machine, and workers electing 45% of corporate boards.
• 3:28
  SPEAKER_00: He has since floated a national moratorium on data centers.
• 3:32
  SPEAKER_00: The head of AFLCIO said that the race to advance AI at all costs with no guardrails is reckless and dangerous.
• 3:41
  SPEAKER_00: So on paper, you have the populist right and the labor left saying the same thing about the same men.
• 3:47
  SPEAKER_00: Bannon calls them broligarchs.
• 3:49
  SPEAKER_00: Bernie calls them oligarchs.
• 3:51
  SPEAKER_00: Same words, same target, different zip codes.
• 3:53
  SPEAKER_00: And meanwhile, who is defending the industry?
• 3:56
  SPEAKER_00: The tech write David Sachs.
• 3:58
  SPEAKER_00: The man the White House put in charge of AI policy says we have to let the private sector cook.
• 4:04
  SPEAKER_00: He calls the safety people a doomer cult.
• 4:07
  SPEAKER_00: He's claimed that there's one billion dollar doomer industry complex out there manufacturing your bad feelings about AI.
• 4:14
  SPEAKER_00: Mark Andreessen wrote an entire manifesto, and it's public.
• 4:18
  SPEAKER_00: You can read it in 10 minutes, and I definitely want you to, because I couldn't make this up.
• 4:23
  SPEAKER_00: He wrote, We have enemies, and then he listed them.
• 4:26
  SPEAKER_00: And the enemies list literally includes the words sustainability, social responsibility, trust and safety, tech ethics, and risk management.
• 4:35
  SPEAKER_00: Risk management is an enemy.
• 4:38
  SPEAKER_00: The precautionary principle is an enemy.
• 4:41
  SPEAKER_00: He wrote that down on purpose and he published it under his own name.
• 4:45
  SPEAKER_00: So sit with that for just a second.
• 4:47
  SPEAKER_00: The party of family values is fighting the technology companies, and the party of the working man is fighting the technology companies.
• 4:54
  SPEAKER_00: And the technology companies are being defended by a guy who wrote down that risk management is the enemy.
• 5:01
  SPEAKER_00: This is not left versus right anymore.
• 5:03
  SPEAKER_00: This is something else entirely, and I will tell you what it is at the end.
• 5:07
  SPEAKER_00: There's a super PAC called Leading the Future.
• 5:10
  SPEAKER_00: It lost last August.
• 5:12
  SPEAKER_00: Its funders include Andreessen Horowitz, the president of OpenAI, the co-founder of Palantir, and Perplexity.
• 5:19
  SPEAKER_00: They claim they have raised$125 million through only about$50 million actually shows up in the federal filings, which, by the way, is its own other story.
• 5:30
  SPEAKER_00: They're not hiding what they are doing.
• 5:32
  SPEAKER_00: The people running it are the same operators who ran the crypto industry, Super PEC.
• 5:37
  SPEAKER_00: And that one spent over$130 million in 2024 and helped take out a sitting senator.
• 5:44
  SPEAKER_00: They're running the identical playbook.
• 5:46
  SPEAKER_00: Their stated goal is a single national framework so America can win the race against China.
• 5:52
  SPEAKER_00: That's their case.
• 5:53
  SPEAKER_00: And it's not a stupid case.
• 5:55
  SPEAKER_00: And I will steel man it for you in a minute.
• 5:58
  SPEAKER_00: Now, here's what is what has happened when they pulled the trigger.
• 6:02
  SPEAKER_00: Alex Bores in New York.
• 6:05
  SPEAKER_00: Boers wrote New York's AI safety law.
• 6:07
  SPEAKER_00: They spent about$8 million to destroy him.
• 6:10
  SPEAKER_00: The safety side came back with about$20 million, all in, outside money, in that one single house primary cross cost$40 million.
• 6:19
  SPEAKER_00: $40 million for one seat in a primary and they won.
• 6:24
  SPEAKER_00: They beat him.
• 6:25
  SPEAKER_00: They replaced him with a man called Michael Lasher, who co-sponsored Alex Boris' AI safety bill.
• 6:31
  SPEAKER_00: They spent$8 million to swap out an AI safety guy for an AI safety guy.
• 6:36
  SPEAKER_00: And Lasher walked out on stage and said, and I'm quoting him, I have some news for the two big AI companies that have taken such an unusual interest on who won this congressional seat.
• 6:48
  SPEAKER_00: I will not be taking my cues from either of you when it comes to protecting our kids, our jobs, and our environment.
• 6:55
  SPEAKER_00: That is the most expensive own goal in modern political history, and it happened three weeks ago.
• 7:00
  SPEAKER_00: Now the other side, they have money too.
• 7:03
  SPEAKER_00: This there's a counter network called Public First that has pledged around 50 million.
• 7:08
  SPEAKER_00: Anthropic put 20 million into it.
• 7:11
  SPEAKER_00: So do not let me paint this as saints against sinners.
• 7:16
  SPEAKER_00: There is money on both sides, and there are agendas on both sides as well.
• 7:20
  SPEAKER_00: But the guy running the pro-regulation side said something that I think explains this entire episode better than anything else I've read.
• 7:27
  SPEAKER_00: He said, quote, I often tell people which of these sides would you rather have?
• 7:33
  SPEAKER_00: We have$50 million and 85% of the public sentiment.
• 7:38
  SPEAKER_00: They have$100 million and 15% of public opinion.
• 7:43
  SPEAKER_00: Which of those is most likely to win?
• 7:46
  SPEAKER_00: Hold that thought because we're about to look at the public.
• 7:49
  SPEAKER_00: And this next one made me laugh out loud.
• 7:51
  SPEAKER_00: The Pro AI Super PAC has a friendly news site in its orbit, launched right after Christmas.
• 7:58
  SPEAKER_00: 94 articles, not one human wrote any of them.
• 8:01
  SPEAKER_00: An AI detection scan found 69% of the content was fully machine generated, and it gets better.
• 8:07
  SPEAKER_00: The bots have names, fake bylines, fake email addresses, and they were emailing real experts, real policy people, asking for real quotes for real articles, pretending to be reporters.
• 8:18
  SPEAKER_00: One of the people who got an email from a robot pretending to be a journalist was the general counsel of an actual AI policy organization.
• 8:27
  SPEAKER_00: The A lobby got caught astroturfing with AI.
• 8:30
  SPEAKER_00: Now I want to be careful and fair here.
• 8:32
  SPEAKER_00: I'm not saying the super PAC ran that site.
• 8:34
  SPEAKER_00: The connection is through a public relations firm whose chief executive co-founded the PAC, the PAC.
• 8:41
  SPEAKER_00: That is the link.
• 8:42
  SPEAKER_00: Go check it yourself.
• 8:43
  SPEAKER_00: But you have to admit, if you're writing a satire of this industry and you pitch the scene where the robots pretend to be reporters to lobby against robot regulation, an editor would tell you it's too close on the nose.
• 8:56
  SPEAKER_00: While we are here, let's do the rest of the comedy because you deserve it.
• 9:00
  SPEAKER_00: Last month, a Republican Congresswoman submitted an amendment summary to the House Rules Committee.
• 9:06
  SPEAKER_00: She left a string in the text that said, and I quote, 1125 AM question mark, question mark, question mark.
• 9:15
  SPEAKER_00: Claude responded, she denied that AI wrote the legislation, said that staff only used it to check the grammar on the summary, said most staff use it, and then unprompted, she volunteered that she loves Claude, but Grok is way more savage, which is a real sentence sent by a member of Congress, a real one.
• 9:36
  SPEAKER_00: And the machines writing our laws, well, it's about them.
• 9:39
  SPEAKER_00: Speaking of Grok, last summer, that chatbot spent several hours generating anti-Semitic content calling himself itself Mechel Hitler, not a typo Mecha Hitler.
• 9:52
  SPEAKER_00: The company said it was an unauthorized modification to an upstream code path.
• 9:56
  SPEAKER_00: And this the statement came around one day after they announced contracts, making their products available to every federal department and agency, including the Department of Defense.
• 10:06
  SPEAKER_00: Then we have the federal health report that cited studies that do not exist, not misquoted, but they don't exist.
• 10:12
  SPEAKER_00: Four of them, ghost papers, in the classic signature of a machine that made something up and a human who didn't check.
• 10:19
  SPEAKER_00: The White House press secretary called them formatting issues.
• 10:23
  SPEAKER_00: And my favorite, Little Dillotil, the giant consulting firm, delivered a report to the Australian government for about$440,000.
• 10:33
  SPEAKER_00: It contained fake academic citations and a completely fabricated quote from a federal court ruling.
• 10:39
  SPEAKER_00: The judge said a thing he never said.
• 10:43
  SPEAKER_00: Academic caught it.
• 10:47
  SPEAKER_00: Delotile refunded part of the money and quietly added a note to the revised version disclosing what they had.
• 10:54
  SPEAKER_00: In fact, used a large language model.
• 10:56
  SPEAKER_00: They build a government 400 grand for a book report that the machine hallucinated.
• 11:02
  SPEAKER_00: All right.
• 11:03
  SPEAKER_00: This is fun.
• 11:03
  SPEAKER_00: Now let's do the hard part because I promised you both sides that I meant it.
• 11:07
  SPEAKER_00: Do not let the political circus convince you that nothing real is happening.
• 11:12
  SPEAKER_00: Something enormous is happening.
• 11:14
  SPEAKER_00: Start with costs.
• 11:15
  SPEAKER_00: In November of 2022, getting a certain level of intelligence out of a machine costs about$20 per million tokens.
• 11:22
  SPEAKER_00: Two years later, same level of capability cost 7 cents.
• 11:26
  SPEAKER_00: $20 to 7 cents.
• 11:28
  SPEAKER_00: Not an improvement, that's a collapse.
• 11:31
  SPEAKER_00: Nothing in the history of the industry has fallen in price that fast.
• 11:34
  SPEAKER_00: Not steel, not oil, not computation itself.
• 11:38
  SPEAKER_00: Now medicine.
• 11:39
  SPEAKER_00: Last January in the Lancet, they published the results of a trial in Sweden with 105,000 women.
• 11:47
  SPEAKER_00: Real randomized trial, not a press release, AI-supported mammogram screening, cut interval cancers.
• 11:53
  SPEAKER_00: The ones that show up between screenings that tend to be the ones that kill you by 12%.
• 11:58
  SPEAKER_00: 27% fewer of the aggressive kind.
• 12:01
  SPEAKER_00: Not a chatbot writing your emails.
• 12:03
  SPEAKER_00: That's not that.
• 12:04
  SPEAKER_00: This is women being alive in 2030 who could not have been before.
• 12:11
  SPEAKER_00: And I'll give you the honest footnote because I'm not going to sell you a headline.
• 12:14
  SPEAKER_00: The lead author of that study said explicitly that it does not support replacing radiologists.
• 12:21
  SPEAKER_00: The AI arm still needs a human.
• 12:24
  SPEAKER_00: If we made the humans better, it did not delete them.
• 12:28
  SPEAKER_00: Which brings me to the single best story in this entire show.
• 12:30
  SPEAKER_00: In 2016, Jeffrey Hinton, the man they call the godfather of artificial intelligence, the man who won the Nobel Prize, told the world to stop training radiologists.
• 12:40
  SPEAKER_00: He said within five years, deep learning would do better.
• 12:43
  SPEAKER_00: Stop training them.
• 12:44
  SPEAKER_00: He did not enter the field.
• 12:46
  SPEAKER_00: That was 10 years ago.
• 12:48
  SPEAKER_00: Now, today the United States has the largest radiologist shortage in its history.
• 12:53
  SPEAKER_00: Radiologists are making around$571,000 in a year, and their pay went up 9% last year.
• 13:01
  SPEAKER_00: Mayo Clinic's radiology staff has grown 55% since Hinton said that.
• 13:07
  SPEAKER_00: He admitted he was wrong.
• 13:08
  SPEAKER_00: He says he was wrong on the timing, not the direction.
• 13:13
  SPEAKER_00: Maybe.
• 13:14
  SPEAKER_00: But about the young person who heard that in 2016 from a Nobel laureate and picked a different career.
• 13:19
  SPEAKER_00: Think about what that advice cost him.
• 13:22
  SPEAKER_00: And then think very carefully about how much confidence you should place in anybody, on any side, telling you exactly what this technology is going to do to your job.
• 13:31
  SPEAKER_00: That's the case for it's real.
• 13:34
  SPEAKER_00: It's the enormous, and it's the most famous profit gap got that most famous prediction specially wrong.
• 13:42
  SPEAKER_00: Now, the case against it, and I'm going to make this one properly too.
• 13:46
  SPEAKER_00: Forget the robot apocalypse.
• 13:48
  SPEAKER_00: The strongest argument against this whole thing is boring and it's on your utility bill.
• 13:53
  SPEAKER_00: There's a power grid in the eastern part of the country called PJM.
• 13:56
  SPEAKER_00: It serves 67 million people.
• 13:59
  SPEAKER_00: It runs an auction to lock in electricity capacity.
• 14:03
  SPEAKER_00: Three auctions ago, the capacity cleared at about$29.
• 14:07
  SPEAKER_00: The most recent one cleared at$333, 11.5 times higher.
• 14:12
  SPEAKER_00: In three years, PJM said in its own release that the increase in the peak lead forecast, 97% of it, is data centers.
• 14:21
  SPEAKER_00: Now here's the part that should light you up.
• 14:23
  SPEAKER_00: The independent market monitor found that$6.5 billion of that auction cost was for data centers, and$6.2 billion of that$6.5 billion for data centers that have not been built yet.
• 14:37
  SPEAKER_00: You're paying right now, today, to reserve power for buildings that don't exist.
• 14:42
  SPEAKER_00: In Virginia, the typical residential bill goes up$11 in change a month this year.
• 14:47
  SPEAKER_00: New Jersey just signed the first data center tariff law in the country.
• 14:51
  SPEAKER_00: Six days ago, Ohio now makes data centers pay 85% of the power they reserve, whether they use it or not, for 12 years.
• 15:00
  SPEAKER_00: And the public has figured it out.
• 15:02
  SPEAKER_00: Gallup, two months ago, 71% of Americans oppose data centers being built in their community.
• 15:08
  SPEAKER_00: 48% strongly oppose it.
• 15:10
  SPEAKER_00: For comparison, only 53% oppose a nuclear power plant.
• 15:15
  SPEAKER_00: Americans would rather have a reactor in their backyard than a building full of computers.
• 15:20
  SPEAKER_00: Now I'm going to do something that the people on my side of the aisle don't do, and I'm going to tell you which of the AI anti-AI arguments is weak, because if I only hand you ammunition that fires, you're going to get hurt even when it jams.
• 15:33
  SPEAKER_00: Now, the water argument's weak.
• 15:35
  SPEAKER_00: You have heard it that AI is draining our water.
• 15:37
  SPEAKER_00: Google used about 6 billion gallons across all of its data centers in a year.
• 15:42
  SPEAKER_00: America golf courses use about 547 billion gallons.
• 15:47
  SPEAKER_00: Golf courses use roughly 70 times more water than Google.
• 15:52
  SPEAKER_00: And your chatbot query uses a quarter of a watt an hour, which is nothing.
• 15:57
  SPEAKER_00: 10 billion chat messages a day would be about 2% of AI's power drop.
• 16:02
  SPEAKER_00: The water in the chatbot stuff isn't where the damage is.
• 16:05
  SPEAKER_00: The damage is in the sitting, where you put the thing and in training and in the big agent workloads.
• 16:11
  SPEAKER_00: So if you go on the internet and yell about water, somebody who knows some real numbers are gonna, they're gonna hand you your head.
• 16:18
  SPEAKER_00: And then they'll be right.
• 16:21
  SPEAKER_00: When then and then they use it and say, dismiss everything else you said.
• 16:25
  SPEAKER_00: So do not do it.
• 16:26
  SPEAKER_00: Attack those strong points, and there are plenty.
• 16:29
  SPEAKER_00: And on jobs, here's the honest version, and both sides are gonna hate it.
• 16:33
  SPEAKER_00: Stanford found a 16% relative employment decline for 22 to 25 year olds in the most AI exposed jobs.
• 16:41
  SPEAKER_00: That's real, and it's scary, and it's the best evidence the pessimists have.
• 16:45
  SPEAKER_00: But the men who wrote that paper, Eric Bejenson, it's difficult to say, sorry, went back and looked again in February, and he said in his own words, and the effect only becomes statistically significant starting in 2024, and the earlier declines were probably caused by other things.
• 17:03
  SPEAKER_00: What other things?
• 17:04
  SPEAKER_00: The Federal Reserve raising interest rates in 2022 before JetGPT even existed.
• 17:10
  SPEAKER_00: The author of the scariest AI jobs paper in America partially walked it back himself in public and almost nobody reported it.
• 17:18
  SPEAKER_00: Meanwhile, Brookings looked at the same landscape and found no relationship at all between AI exposure and unemployment.
• 17:25
  SPEAKER_00: And unemployment right now is 4.2%.
• 17:28
  SPEAKER_00: Not the bloodbath, not yet.
• 17:31
  SPEAKER_00: So what's actually true?
• 17:32
  SPEAKER_00: Entry level is getting squeezed.
• 17:34
  SPEAKER_00: Seniors are fine.
• 17:36
  SPEAKER_00: Companies love saying AI did it because it makes them sound like the future instead of sounding like the overhead in 2021.
• 17:43
  SPEAKER_00: And a young person trying to get in the door is having a harder time than at any point in my lifetime.
• 17:49
  SPEAKER_00: That's not the apocalypse, but if you're if you're 23, it's your apocalypse.
• 17:54
  SPEAKER_00: Both things are true.
• 17:56
  SPEAKER_00: Now the segment I think matters most, the receipts.
• 17:58
  SPEAKER_00: Everybody in the fight makes predictions.
• 18:00
  SPEAKER_00: Nobody keeps score.
• 18:02
  SPEAKER_00: So today we're going to keep score.
• 18:03
  SPEAKER_00: In March of last year, the chief executive of Anthropic said AI would be writing 90% of all code within three to six months.
• 18:10
  SPEAKER_00: Essentially all of it within a year.
• 18:12
  SPEAKER_00: That was 16 months ago.
• 18:14
  SPEAKER_00: I write code.
• 18:15
  SPEAKER_00: It's not writing 90% of it.
• 18:17
  SPEAKER_00: His own company surveyed its own engineers and found roughly two times speed up.
• 18:21
  SPEAKER_00: Two times is incredible.
• 18:23
  SPEAKER_00: Two times is not 90%, though.
• 18:25
  SPEAKER_00: Elon Musk predicted artificial intelligence by 2025.
• 18:29
  SPEAKER_00: Then he predicted it by 2026.
• 18:31
  SPEAKER_00: At Davos in January, he moved it again.
• 18:34
  SPEAKER_00: To the end of this year, that prediction has a rolling 12-month horizon, and it now has rolled three times.
• 18:40
  SPEAKER_00: Last October, an executive at OpenAI announced their model had solved 10 previously unsolved math problems.
• 18:47
  SPEAKER_00: That's huge.
• 18:48
  SPEAKER_00: That's historic.
• 18:49
  SPEAKER_00: Machine does original mathematics.
• 18:52
  SPEAKER_00: It was retracted in about 17 hours.
• 18:54
  SPEAKER_00: The man who runs the website where those problems live said it was a dramatic misrepresentation.
• 19:00
  SPEAKER_00: The problems were listed as open only because he personally had not seen the paper solving them.
• 19:06
  SPEAKER_00: The model had found the existing papers.
• 19:08
  SPEAKER_00: It did a beautiful job of looking things up, and a senior executive at the leading AI company on Earth announced it as a mathematical discovery.
• 19:18
  SPEAKER_00: The chief executive of DeepMind said out loud, This is embarrassing.
• 19:21
  SPEAKER_00: Now, turn the gun around because the Doomers have a scorecard too.
• 19:26
  SPEAKER_00: Remember when 2024 was going to be the year Deep Fakes broke democracy?
• 19:31
  SPEAKER_00: Every outlet ran it.
• 19:32
  SPEAKER_00: Harvard's Ash Center went and checked afterward.
• 19:35
  SPEAKER_00: Their headline was the apocalypse that wasn't.
• 19:38
  SPEAKER_00: Researchers found that low-tech fakes, the dumb kind, the miscaptioned photo, were used seven times more often than AI generated content.
• 19:47
  SPEAKER_00: Meta reported that less than 1% of all fact-checked mess information in the 2024 election cycles was AI.
• 19:55
  SPEAKER_00: Well, there's now actual research paper titled, We Built Detectors for a Threat That Didn't Arrive.
• 20:01
  SPEAKER_00: And the most honest thing in this entire show, and I have real respect for this, comes from the Doomers themselves.
• 20:06
  SPEAKER_00: The authors of the most influential AI doom forecast in the world went back this year and graded their own homework in public, their finding.
• 20:15
  SPEAKER_00: Reality is running about 65 to 70% of the pace they predicted.
• 20:19
  SPEAKER_00: Their benchmark predictions were too aggressive.
• 20:22
  SPEAKER_00: One author's timeline for AGI slipped from 2028 to 2029.
• 20:28
  SPEAKER_00: The other one now says early 2030s.
• 20:32
  SPEAKER_00: Nobody made them do that.
• 20:34
  SPEAKER_00: They just did that.
• 20:35
  SPEAKER_00: And this is more intellectual honesty than I've seen in from a billion dollar lab or any other super PAC in this entire fight.
• 20:42
  SPEAKER_00: Now let me turn this on myself because I have to do this part or the last 20 minutes were a sales pitch.
• 20:47
  SPEAKER_00: I use this technology every day.
• 20:49
  SPEAKER_00: It's the reason a one-man operation could do the work of a team.
• 20:52
  SPEAKER_00: It has genuinely changed my life.
• 20:54
  SPEAKER_00: So when I tell you the industry is behaving badly, understand that I'm also a customer and I'm a happy one.
• 21:01
  SPEAKER_00: And you should discount me accordingly.
• 21:04
  SPEAKER_00: I'll give you the accelerationist their strongest card because it's a real card.
• 21:08
  SPEAKER_00: If America slams the brakes and China doesn't, we do not get a safer world.
• 21:13
  SPEAKER_00: We get that same world with a different owner.
• 21:16
  SPEAKER_00: That's not a stupid argument, that's a serious argument, and the people making it are not all shills.
• 21:22
  SPEAKER_00: Some of them genuinely believe that the fastest path to safety runs through winning.
• 21:28
  SPEAKER_00: I don't know if I fully agree, but I refuse to pretend it's dumb.
• 21:32
  SPEAKER_00: And here's the one that cuts me.
• 21:34
  SPEAKER_00: The public is scared of this thing, and the public has been scared of everything.
• 21:38
  SPEAKER_00: They were scared of the automobile.
• 21:40
  SPEAKER_00: They're scared of electricity in the home.
• 21:42
  SPEAKER_00: 80% of Americans telling Gallup they would rather slow AI down than lose the safety rules is a real number, and it deserves respect.
• 21:50
  SPEAKER_00: But a majority has been wrong before, and being popular is not the same thing as being right.
• 21:56
  SPEAKER_00: So hold on to all of that at once.
• 21:59
  SPEAKER_00: That is the job.
• 22:01
  SPEAKER_00: Here's what I think is actually going on, and it's not left versus right.
• 22:05
  SPEAKER_00: It's old fight regulation versus deregulation.
• 22:09
  SPEAKER_00: That fight's over.
• 22:10
  SPEAKER_00: It died last summer, 99-1.
• 22:13
  SPEAKER_00: The new fight between the people who own the machines and everybody else, well, that line doesn't run between all the parties.
• 22:20
  SPEAKER_00: It runs straight through the middle of both of them.
• 22:23
  SPEAKER_00: And this is why a man like Bannon and a man like Bernie reach for the same word.
• 22:28
  SPEAKER_00: This is why 99 senators voted together, and this is why Republicans and Democrats are now within one point of each other on how worried they are.
• 22:36
  SPEAKER_00: 50% and 51%, which basically never happens in this country about anything.
• 22:42
  SPEAKER_00: And it's why an industry with$125 million in the White House is losing the argument 80 to 9 and has to spend that money buying it back.
• 22:53
  SPEAKER_00: The tech writes a fissile explanation for why you do not like this is that a billion-dollar Doomer conspiracy brainwashed you.
• 23:02
  SPEAKER_00: Gallup's explanation is that you do not want a warehouse full of computers behind your house raising your power bill.
• 23:09
  SPEAKER_00: I think Gallup's closer, but here's the thing I keep coming back to, and it's why I'm not a doomer and will never be.
• 23:15
  SPEAKER_00: The technology is not the villain in the story.
• 23:18
  SPEAKER_00: The technology is the most powerful lever a working person has ever been handed.
• 23:22
  SPEAKER_00: One person working with these tools can do what a company did of 20 did when I started.
• 23:28
  SPEAKER_00: That's not a threat to the little guy.
• 23:30
  SPEAKER_00: That's the little guy's shot.
• 23:32
  SPEAKER_00: The threat is the lever gets locked in for building that you pay electric bill for and never get to walk into.
• 23:38
  SPEAKER_00: So we watch the money.
• 23:39
  SPEAKER_00: We watch who writes the rules and who paid for the pen.
• 23:43
  SPEAKER_00: We refuse to pick a team just because a jersey got handed to us.
• 23:47
  SPEAKER_00: And we keep asking the only question that actually matters: who owns the machine?
• 23:51
  SPEAKER_00: Who does it answer to?
• 23:52
  SPEAKER_00: Because if the answer is you, this is the greatest thing that ever happened to a person working alone.
• 23:57
  SPEAKER_00: And if the person is somebody else, then it doesn't matter how smart it gets.
• 24:02
  SPEAKER_00: It was never yours to begin with.
• 24:04
  SPEAKER_00: Same time tomorrow.
• 24:05
  SPEAKER_00: We keep watch together.
• 24:06
  SPEAKER_00: Be well, be safe, and take care of yourselves.
• 24:09
  SPEAKER_00: I'm Connor with Honor.
• 24:10
  SPEAKER_00: Over and out.