Latest / Connor with Honor: AI, Real Estate and Real Life / The Respectable Word: Only 3,462 August Job Cuts Were Blamed on AI

Transcript

The Respectable Word: Only 3,462 August Job Cuts Were Blamed on AI · Connor with Honor: AI, Real Estate and Real Life

• 0:00
  SPEAKER_00: Somebody in your life read a news article this week that was written by a machine on a website that doesn't employ a single human being.
• 0:07
  SPEAKER_00: And you know they believed it.
• 0:08
  SPEAKER_00: And I know you got those people like I got those people.
• 0:11
  SPEAKER_00: They believe it.
• 0:12
  SPEAKER_00: And then they told you about it at work.
• 0:14
  SPEAKER_00: And there were about 70 of those websites inside of one operation.
• 0:18
  SPEAKER_00: 70 fabricated news sites plus more than 250 inauthentic commenting accounts underneath them, running across six continents, sold as a service to whomever paid the bill.
• 0:29
  SPEAKER_00: And that's not a prediction and it's not a movie.
• 0:32
  SPEAKER_00: It's mass disinformation used to create unease across every platform, every societal structure, everywhere at the same time, paid by people who apparently will remain nameless, at least for now.
• 0:50
  SPEAKER_00: And it was in a report published yesterday by the company whose own model was being used to do it.
• 0:57
  SPEAKER_00: You see these large language models, they're really powerful.
• 1:00
  SPEAKER_00: They are not only able to potentially, as long as the failsafes are there, they can't, but they're capable, if left unlocked, to seek out biological destruction.
• 1:12
  SPEAKER_00: They could do that.
• 1:13
  SPEAKER_00: Well, that's not the only thing they could do.
• 1:15
  SPEAKER_00: They can master a mega media campaign and put, if you have the money, put 10,000 agents with all PhD level, level educations and misinformation and societal manipulation.
• 1:34
  SPEAKER_00: And you know, we're easy.
• 1:37
  SPEAKER_00: I like the little cats that dance around and that uh that really upbeat music.
• 1:41
  SPEAKER_00: I like that.
• 1:42
  SPEAKER_00: That's easy.
• 1:43
  SPEAKER_00: Okay, I want you to hold a number for me.
• 1:45
  SPEAKER_00: 3,462.
• 1:46
  SPEAKER_00: I'm gonna come back to it in about nine minutes.
• 1:48
  SPEAKER_00: And when I do, I think some weight off your chest.
• 1:52
  SPEAKER_00: I think it'll take a little bit.
• 1:53
  SPEAKER_00: Now, you've been carrying around all year this weight on your chest.
• 1:56
  SPEAKER_00: You haven't been asked to carry it around, so I'll get to it.
• 1:59
  SPEAKER_00: Good morning, good afternoon, good evening, of course.
• 2:00
  SPEAKER_00: Whenever you happen to be tuning in, this is AI with honor, the daily download.
• 2:04
  SPEAKER_00: It lives at Santa Clarita Artificial Intelligence, the practical AI porter for Santa Clarita business owners.
• 2:10
  SPEAKER_00: Everything in this read came off a primary source.
• 2:13
  SPEAKER_00: So if a company says something about its own product, I'm gonna tell you it's a company claim.
• 2:17
  SPEAKER_00: If a number came from the government, I'm gonna tell you that too.
• 2:20
  SPEAKER_00: There are different kinds of things, and the difference is most of the argument.
• 2:25
  SPEAKER_00: And it's all very self-interest driven.
• 2:27
  SPEAKER_00: I mean, I don't know how many folks are out there without trying to drive some kind of an agenda.
• 2:34
  SPEAKER_00: I don't monetize this.
• 2:36
  SPEAKER_00: I there's no way.
• 2:37
  SPEAKER_00: I mean, three followers, one watch, I think it's me reviewing it when I'm done.
• 2:42
  SPEAKER_00: So don't like it, don't subscribe, don't do anything.
• 2:45
  SPEAKER_00: I don't want to get caught up in all that mess.
• 2:46
  SPEAKER_00: But please comment if that helps.
• 2:49
  SPEAKER_00: I'm not trying to help the algorithm, but I would love to at least get a conversation going with you.
• 2:53
  SPEAKER_00: But you're not required to do anything.
• 2:55
  SPEAKER_00: I didn't talk to my employees and and they keep telling me something about my particular where most of you got unsubscribed.
• 3:02
  SPEAKER_00: Yeah, nothing like that.
• 3:03
  SPEAKER_00: Just just I hope it helps.
• 3:05
  SPEAKER_00: I hope it helps.
• 3:06
  SPEAKER_00: And don't believe anything.
• 3:08
  SPEAKER_00: You gotta really look into it.
• 3:10
  SPEAKER_00: You gotta do a little bit more heavy lifting these days.
• 3:13
  SPEAKER_00: And the system, you go to look it up, it might be lying like the original source.
• 3:18
  SPEAKER_00: All right, let's get back to this.
• 3:20
  SPEAKER_00: Anthropic published its threat intelligence report yesterday, September 10th.
• 3:24
  SPEAKER_00: It covers what their team found.
• 3:26
  SPEAKER_00: We're 2026, by the way.
• 3:27
  SPEAKER_00: Just God, I hate that.
• 3:29
  SPEAKER_00: Don't you?
• 3:29
  SPEAKER_00: You're watching that video and you think, God, this sounds kind of sounds familiar.
• 3:32
  SPEAKER_00: And you look at say a year ago, and they just republicized it or published it.
• 3:36
  SPEAKER_00: Yeah, that's also sucks.
• 3:38
  SPEAKER_00: Okay.
• 3:40
  SPEAKER_00: It covers what their team found and shut down between December 2025 and August 2026 across seven categories: cyber operations, surveillance, influence, sorry about that, influence operations.
• 3:52
  SPEAKER_00: I forgot I had it.
• 3:53
  SPEAKER_00: Conventional weapons, biological misuse, scams and fraud, and distillation, like the water, but different, which is stealing a model by asking it a very large number of questions and training your own thing on the answers.
• 4:07
  SPEAKER_00: Distillation.
• 4:08
  SPEAKER_00: Yeah, I guess they all kind of do it.
• 4:11
  SPEAKER_00: What's the word?
• 4:12
  SPEAKER_00: Hasn't been proven.
• 4:13
  SPEAKER_00: I forgot what the word is, real quick.
• 4:15
  SPEAKER_00: All right.
• 4:16
  SPEAKER_00: The influence, uh, the influence operation I opened with is one case of mini out there.
• 4:21
  SPEAKER_00: Roughly 70 fabricated news websites, more than 250 inauthenticamenting accounts, run as a paid commercial service across six continents.
• 4:31
  SPEAKER_00: Somebody sent an invoice for that.
• 4:32
  SPEAKER_00: Somebody paid it.
• 4:33
  SPEAKER_00: Label this one carefully with me.
• 4:35
  SPEAKER_00: This is a company reporting on misuse of its own product, and that's not a reason to throw it out.
• 4:40
  SPEAKER_00: It's is a reason to notice that they hold two incentives at once, though, to show you that they're catching things.
• 4:45
  SPEAKER_00: Good job.
• 4:46
  SPEAKER_00: And to show you that the thing is powerful.
• 4:49
  SPEAKER_00: Good job.
• 4:51
  SPEAKER_00: We all like gangster.
• 4:52
  SPEAKER_00: Come on.
• 4:53
  SPEAKER_00: Let's let's get over this.
• 4:54
  SPEAKER_00: Gotta be careful though.
• 4:55
  SPEAKER_00: Now, both can be true, can be true while the findings are also true.
• 5:00
  SPEAKER_00: The line in that report that actually matters is not the fake news count, it is this one.
• 5:05
  SPEAKER_00: Anthropic says the model's role in cyber operations has become, in their words, increasingly autonomous with multi-agent frameworks running reconnaissance, exploitation, and data exfiltration, and humans staying in the loop mainly to set the targets.
• 5:20
  SPEAKER_00: Read that slowly.
• 5:21
  SPEAKER_00: Humans mainly in the loop to set the targets because it's the description of a workflow, not a monster.
• 5:26
  SPEAKER_00: Person picks up a goal.
• 5:28
  SPEAKER_00: The software surveys the ground, writes its own script, runs them, reads what came back, and goes again until the job has completed.
• 5:36
  SPEAKER_00: Now here's the uncomfortable part, and it's uncomfortable for the sales pitch, not for you.
• 5:41
  SPEAKER_00: This is the identical loop being sold to your plumbing company as an agent.
• 5:46
  SPEAKER_00: Same architecture, same autonomy.
• 5:48
  SPEAKER_00: The only difference in the whole picture is what the human pointed it at.
• 5:52
  SPEAKER_00: Same day, September 10th.
• 5:53
  SPEAKER_00: OpenAI shipped two things.
• 5:55
  SPEAKER_00: The first is an agent's API with a company says let's developer build agents reliable enough to put into production.
• 6:01
  SPEAKER_00: This is a company claim, and the company's own words and the reliability is precisely what has been missing from every agent demo for two years.
• 6:10
  SPEAKER_00: It's what you watch for now, somebody outside the company measuring it.
• 6:14
  SPEAKER_00: Now, the second's more interesting to a business owner.
• 6:17
  SPEAKER_00: ChatGPT for financial services, it pairs their GPT-6 Astro model, which is, yeah, gotta say, I wasn't a big fan at first, but yeah, I get it.
• 6:28
  SPEAKER_00: Astro model with the licensed financial data, but with Morgan Stanley and Evercore's design partners using data sets from Dalupa, plus Pitchbook, L S E G news, and Crunch Base.
• 6:41
  SPEAKER_00: OpenAI hosts and indexes that data itself so the thing can cite a specific table on a specific page instead of telling you a beautiful story.
• 6:51
  SPEAKER_00: And read the boring part because the boring part is the actual product in this case.
• 6:55
  SPEAKER_00: Business data is not used to train models by default, it's encrypted at rest and in transit.
• 7:02
  SPEAKER_00: Administrators can figure retention.
• 7:04
  SPEAKER_00: Compliance teams can export audit logs, look at the shape of that thing because the shape is the whole lesson, and the shape is free.
• 7:12
  SPEAKER_00: They did not build something that knows everything.
• 7:14
  SPEAKER_00: They picked one narrow job, investment, banking, and equity research.
• 7:18
  SPEAKER_00: They licensed real data instead of hoping the model remembered it correctly.
• 7:22
  SPEAKER_00: They made it cite the exact table it pulled from and they put retention and access controls in the box while they have sold it to anybody before they did that.
• 7:32
  SPEAKER_00: Narrow job, real data citations controls.
• 7:35
  SPEAKER_00: That's what development looks like.
• 7:36
  SPEAKER_00: And it would be awesome if we could take these instead of trying to jump into this general intelligence, which might be okay, but maybe needs to be a little bit safer.
• 7:45
  SPEAKER_00: We just poke these things, we focus them right on all the cancer research, all of the disease, all of the longevity problems, all of the aging issues, every kind of sickness out there, and world peace, and oh my God, but no, no.
• 8:01
  SPEAKER_00: So narrow job, real data, citations, controls, that's what deployment looks like.
• 8:05
  SPEAKER_00: Now, everything else you've been shown this year is a demo.
• 8:08
  SPEAKER_00: You don't need Morgan Stanley budget to copy that list.
• 8:11
  SPEAKER_00: You need to pick one job.
• 8:12
  SPEAKER_00: The list is the expensive part, and somebody just published it for free.
• 8:16
  SPEAKER_00: 3,462.
• 8:18
  SPEAKER_00: That's how many announced job cuts in the United States in the month of August were attributed to artificial intelligence.
• 8:24
  SPEAKER_00: According to Challenger, Gray and Christmas, the firm that has been counting announced layoffs in this country since the 1990s, 3,462 in a month that had 52,881 announced cuts in total.
• 8:38
  SPEAKER_00: It's the lowest monthly AI figure since December of 2025.
• 8:42
  SPEAKER_00: My question to all of those people that were laid off, if it really is true that they were laid off because of AI, isn't AI coming into the business?
• 8:50
  SPEAKER_00: Isn't it going to make that person's position better?
• 8:54
  SPEAKER_00: And isn't it going to make more money for the company than that person was capable of?
• 8:59
  SPEAKER_00: Now, if the human being is the least able person on the team, why don't they just keep the human?
• 9:07
  SPEAKER_00: AI is working, AI's doing all this fancy stuff.
• 9:10
  SPEAKER_00: Just keep the human.
• 9:11
  SPEAKER_00: AI is going to double their triple their production at least.
• 9:15
  SPEAKER_00: That's what it would be.
• 9:17
  SPEAKER_00: And humans eight hours.
• 9:18
  SPEAKER_00: AI got a couple more of those shifts, 24 hours a day, seven days a week, no days off, no vacation, no sick time, no nothing.
• 9:26
  SPEAKER_00: And it's there and it's working.
• 9:28
  SPEAKER_00: Let's keep them, folks.
• 9:29
  SPEAKER_00: Come on, talk to somebody.
• 9:31
  SPEAKER_00: We shouldn't be getting rid of people for AI because AI is that's the savior, right?
• 9:36
  SPEAKER_00: That's the one that fixes everything.
• 9:38
  SPEAKER_00: So we'll just keep the employee.
• 9:40
  SPEAKER_00: No, they don't have to keep their hands in the cookie jar.
• 9:44
  SPEAKER_00: It's okay.
• 9:44
  SPEAKER_00: You know, have them do something else.
• 9:46
  SPEAKER_00: Have them, I don't know, wash the car.
• 9:49
  SPEAKER_00: I don't know what the other thing would be.
• 9:51
  SPEAKER_00: You want me to figure it all out?
• 9:52
  SPEAKER_00: No, but seriously, though, keep them there.
• 9:55
  SPEAKER_00: Maybe, maybe there's something they can do with AI without creating the biases with it.
• 10:00
  SPEAKER_00: Maybe they could do something without slowing it down or interfering.
• 10:03
  SPEAKER_00: Maybe, maybe there's some additional attribute that the human being can do.
• 10:06
  SPEAKER_00: There's a lot of things.
• 10:07
  SPEAKER_00: I'm sure it would be awesome.
• 10:08
  SPEAKER_00: But again, so we got that unemployment.
• 10:11
  SPEAKER_00: Let me get back to it.
• 10:12
  SPEAKER_00: And it ends a five-month run.
• 10:13
  SPEAKER_00: This was the lowest monthly AI figure since December 2025 when it was 142, and it ends a five-month run, starting back in March, where artificial intelligence was the single most cited reason every month.
• 10:26
  SPEAKER_00: In August, it fell all the way to fourth.
• 10:29
  SPEAKER_00: Here's what beat it.
• 10:30
  SPEAKER_00: Restructuring beat it.
• 10:31
  SPEAKER_00: 16,173 cuts.
• 10:34
  SPEAKER_00: Market and economic conditions, 15,260.
• 10:38
  SPEAKER_00: Closings, 6,743.
• 10:40
  SPEAKER_00: Then artificial intelligence at 3,462.
• 10:44
  SPEAKER_00: 3,462 out of 52,881 is about 6.5%.
• 10:50
  SPEAKER_00: Now, here's why I wanted you to hold that number.
• 10:52
  SPEAKER_00: Search the words AI layoffs this morning and you will find sites telling you that 54% of 2026 layoffs are AI-driven, that 170,000 people lost their jobs to it.
• 11:05
  SPEAKER_00: Go to the film, actually doing the counting and the year-to-date figures.
• 11:09
  SPEAKER_00: 116,175 out of 529,914.
• 11:15
  SPEAKER_00: That's about 22%.
• 11:16
  SPEAKER_00: And it's still the number one cited reason for this year.
• 11:19
  SPEAKER_00: So this is genuinely not nothing, but it's 22%, not 54%.
• 11:24
  SPEAKER_00: And last month it was six and a half.
• 11:27
  SPEAKER_00: Now, I'm going to call this a respectable word.
• 11:29
  SPEAKER_00: When a company has to cut people, artificial intelligence is the most respectable word available to it.
• 11:36
  SPEAKER_00: It says we're advancing.
• 11:38
  SPEAKER_00: Restructuring says a committee got it wrong.
• 11:40
  SPEAKER_00: Market conditions said we misread the year.
• 11:43
  SPEAKER_00: AI says we are the future.
• 11:45
  SPEAKER_00: So in one of those memos lands, run the three possibilities before you panic.
• 11:50
  SPEAKER_00: The machine genuinely does the work now.
• 11:52
  SPEAKER_00: Well, they say it does.
• 11:54
  SPEAKER_00: And if they're replacing human beings with artificial intelligence, yeah, it's got to be at least as good as the human.
• 12:01
  SPEAKER_00: At least as good as the human.
• 12:02
  SPEAKER_00: So keep the human.
• 12:04
  SPEAKER_00: Well, it's free.
• 12:05
  SPEAKER_00: I mean, if you get, I mean, it should be free, right?
• 12:08
  SPEAKER_00: I mean, that's where this is.
• 12:10
  SPEAKER_00: And if it's not free, the extra that it's going to be making because it is super intelligent, so awesome and magical, hell, keep the human.
• 12:18
  SPEAKER_00: I don't see why we should get rid of them.
• 12:21
  SPEAKER_00: Or the company needed to cut anyway, and AI is the polite label on the door.
• 12:24
  SPEAKER_00: Or they're cutting people to pay for the AI because the hardware's not free.
• 12:29
  SPEAKER_00: Well, it's going to be much more productive.
• 12:31
  SPEAKER_00: It gets a two extra shifts.
• 12:33
  SPEAKER_00: You can at least keep the human and pay for the AI because it gets to work 24 hours, not eight, and every day.
• 12:39
  SPEAKER_00: All three are happening in real companies, right?
• 12:41
  SPEAKER_00: Now only one of them is the story you get told and take the rest of this counterweight because I'm sitting here.
• 12:46
  SPEAKER_00: I don't want you to hand it to you in a half picture.
• 12:50
  SPEAKER_00: Announced cuts in the first eight months of 2026 total 529,914.
• 12:56
  SPEAKER_00: That's down 41% against the same eight months of 2025, which ran 892,362.
• 13:04
  SPEAKER_00: Announced hiring plans set at 119,825 for the year.
• 13:09
  SPEAKER_00: That's up 37% over last year.
• 13:12
  SPEAKER_00: And Andy Challenger's own caution on that.
• 13:14
  SPEAKER_00: In his words, while companies are making plans to hire more workers than last year, it doesn't appear those positions are being filled quickly.
• 13:22
  SPEAKER_00: So put the three together.
• 13:23
  SPEAKER_00: Fewer cuts than last year, move more hiring announced than last year, and the jobs are slow to actually get filled.
• 13:31
  SPEAKER_00: That is a strange, sluggish, frustrating labor market, and it's not a collapsing one, and it does not fit either sales pitch that you're being handed.
• 13:39
  SPEAKER_00: Which brings me to the thing I say every day and will keep saying until it stops being true.
• 13:44
  SPEAKER_00: Two groups are selling you fear and they're selling it in opposite directions.
• 13:48
  SPEAKER_00: The lab side says it's escaping.
• 13:50
  SPEAKER_00: It's smarter than you.
• 13:51
  SPEAKER_00: Everything is about to change by the plan.
• 13:54
  SPEAKER_00: The pulpit and the podcast and the campaign side say it's coming for your kids, your job, and your country, your safety and your health.
• 14:01
  SPEAKER_00: Subscribe, donate, and please vote.
• 14:04
  SPEAKER_00: Frightened people are going looking for a savior, and saviors get paid.
• 14:08
  SPEAKER_00: So name the incentive every single time.
• 14:11
  SPEAKER_00: And not to be cynical, but just to be accurate.
• 14:13
  SPEAKER_00: My God, I said it earlier.
• 14:15
  SPEAKER_00: Misinformation, there's a whole campaign, and there's many of them everywhere.
• 14:20
  SPEAKER_00: The threat report I just read you is real work.
• 14:23
  SPEAKER_00: And I believe the findings.
• 14:24
  SPEAKER_00: And the company that published it also sells the model that did those things, and both at the same time, and underneath all of it is the part almost nobody wants.
• 14:33
  SPEAKER_00: We built this.
• 14:34
  SPEAKER_00: There's no outside force and there is no arrival.
• 14:37
  SPEAKER_00: Every item in the report happened because a human being decided to hand a piece of agency to software and point it at something.
• 14:46
  SPEAKER_00: It's a decision made by the people, and decisions can be made differently tomorrow.
• 14:51
  SPEAKER_00: So let's be more useful about that word, agent, because right now it's doing too much work and it's costing people money.
• 14:59
  SPEAKER_00: In your shop, there are three completely different things, and they all get sold under one name.
• 15:03
  SPEAKER_00: There's a switch, something turns it on and off.
• 15:06
  SPEAKER_00: Every new lead gets a text inside 60 seconds.
• 15:09
  SPEAKER_00: That's a switch.
• 15:10
  SPEAKER_00: It requires no intelligence whatsoever, and it will out-earn most of what you're being pitched this quarter.
• 15:16
  SPEAKER_00: And there's a lever, a rule with a condition attached.
• 15:20
  SPEAKER_00: If the lead came out off the yard sign, route it here.
• 15:23
  SPEAKER_00: If it came off the website form after 7 p.m., route it over there.
• 15:27
  SPEAKER_00: Still, just a rule.
• 15:28
  SPEAKER_00: Somebody wrote down once.
• 15:30
  SPEAKER_00: And then there's an agent, the thing out of that threat report.
• 15:33
  SPEAKER_00: You give it a goal and it decides the steps.
• 15:35
  SPEAKER_00: An agent belongs in exactly one place where actual judgment is required.
• 15:40
  SPEAKER_00: And only after you've written down your philosophy, your positioning, and the lines it's never allowed to cross.
• 15:46
  SPEAKER_00: And the more supervised and then supervised it for 30 days because of what leaks in small business doesn't need judgment, needs a switch.
• 15:55
  SPEAKER_00: Now let me make that concrete a little bit.
• 15:57
  SPEAKER_00: Let me weld some iron on it so we can understand the shape of this thing.
• 16:01
  SPEAKER_00: Not one specific client, all right?
• 16:04
  SPEAKER_00: A sick, uh, a sick, six truck heating and air conditioning company.
• 16:09
  SPEAKER_00: Leads arrive in three different ways.
• 16:11
  SPEAKER_00: The phone, a website form, and let's say Google.
• 16:14
  SPEAKER_00: The owner's certain the problem is lead volume, because that's what the entire advertising industry is paid to tell him.
• 16:20
  SPEAKER_00: So, so walk it, walk it down with me.
• 16:23
  SPEAKER_00: The phone rings about 40 times a week, and nine of those go to voicemail between 11:30 and 1 because that's lunchtime.
• 16:30
  SPEAKER_00: And there's one person at the desk.
• 16:31
  SPEAKER_00: The website goes from an inbox somebody checks twice a day.
• 16:35
  SPEAKER_00: Now, the Google messages have never been opened by anyone.
• 16:38
  SPEAKER_00: That's not a volume problem.
• 16:40
  SPEAKER_00: That's nine calls a week leaking out of a business that already paid good money to make that phone ring.
• 16:45
  SPEAKER_00: The fix for the voicemail is a person or a phone system.
• 16:49
  SPEAKER_00: The fix for the form is an automatic text inside of 60 seconds, which is a switch.
• 16:55
  SPEAKER_00: The fix for Google is logging in.
• 16:57
  SPEAKER_00: Total intelligence required zero.
• 17:00
  SPEAKER_00: And that same owner was quoted$2,000 a month for an agent that would have sat on top of all three leaks and answered more slowly than a text message.
• 17:09
  SPEAKER_00: Now, what your own business this weekend?
• 17:12
  SPEAKER_00: Four questions.
• 17:14
  SPEAKER_00: How does the lead come in?
• 17:16
  SPEAKER_00: What happens in the first 10 minutes?
• 17:17
  SPEAKER_00: Who touches it?
• 17:18
  SPEAKER_00: Where does it die?
• 17:20
  SPEAKER_00: One of the standing questions I get every week should I run my own model on my own hardware?
• 17:24
  SPEAKER_00: So nothing I ever type leaves the building?
• 17:27
  SPEAKER_00: Straight answer, you can.
• 17:29
  SPEAKER_00: The hardware to run a genuinely useful model starts in the low thousands of dollars and climbs pretty quickly.
• 17:35
  SPEAKER_00: It pulls real electricity and somebody has to own it, update it, and fix it at 11 p.m.
• 17:40
  SPEAKER_00: on a Thursday when it stops answering.
• 17:42
  SPEAKER_00: What you get back is that nothing you type leaves your building.
• 17:45
  SPEAKER_00: For medical, legal, anything holding a real client file, that's worth the entire price by itself.
• 17:51
  SPEAKER_00: For most shops, it's not worth it yet.
• 17:54
  SPEAKER_00: And the person telling you it definitely is or definitely is not, is usually selling one of those two things.
• 18:00
  SPEAKER_00: The real first question is not the hardware, it's how sensitive the information you're typing in actually is.
• 18:06
  SPEAKER_00: Answer that plainly in the hardware question.
• 18:08
  SPEAKER_00: Well, that answers itself right behind it.
• 18:11
  SPEAKER_00: Two guardrails before I let you go, and both of them are free.
• 18:14
  SPEAKER_00: First, retention.
• 18:15
  SPEAKER_00: Every one of these weapons has a written policy about what happens to the words you type.
• 18:20
  SPEAKER_00: OpenAI puts theirs right in the product announcement.
• 18:23
  SPEAKER_00: They did it yesterday, and the reason they did this is that enterprise buyers demanded it in writing before they would sign.
• 18:30
  SPEAKER_00: You're not an enterprise buyer potentially, so nobody's going to hand you that sentence.
• 18:34
  SPEAKER_00: Go find it yourself.
• 18:35
  SPEAKER_00: If you can't state your vendor's retention policy in one sentence, then you don't actually know where your customer list went.
• 18:42
  SPEAKER_00: Second is flattery.
• 18:44
  SPEAKER_00: These models will tell you your idea is lovely.
• 18:47
  SPEAKER_00: They will tell everybody that.
• 18:49
  SPEAKER_00: Sorry, take to break it to you.
• 18:51
  SPEAKER_00: It's a product decision made to keep your typing, not a verdict on your business.
• 18:55
  SPEAKER_00: If the thing is never once told you no, it's not advising you.
• 19:00
  SPEAKER_00: It's agreeing with you.
• 19:02
  SPEAKER_00: And you can get agreement for free from your brother-in-law.
• 19:05
  SPEAKER_00: So here's one move this week, and it takes about 15 minutes.
• 19:09
  SPEAKER_00: Open the settings of whatever artificial intelligence you use for work.
• 19:13
  SPEAKER_00: Turn off training on the business account if there's a switch for it.
• 19:16
  SPEAKER_00: And on most of them there is.
• 19:17
  SPEAKER_00: Find the retention answer and write it down in one sentence on a paper where you can hand it to somebody.
• 19:22
  SPEAKER_00: Then on that same piece of paper, write the three things you are never going to paste into it: client financials, anything carrying a social security number, anything a competitor would pay money for.
• 19:34
  SPEAKER_00: And even the large language model system themselves, what they would take.
• 19:38
  SPEAKER_00: Now that's the whole agreement.
• 19:40
  SPEAKER_00: 15 minutes, the assignment.
• 19:41
  SPEAKER_00: 15 minutes,$0.
• 19:43
  SPEAKER_00: And when you're done, you'll have performed more actual governance than most people with a policy binder on a shelf.
• 19:49
  SPEAKER_00: And look forward without pretending to see the future because nobody can, and the people charging for the view or the tell.
• 19:56
  SPEAKER_00: If this thing eventually solves everything, the real question becomes who can afford everything.
• 20:01
  SPEAKER_00: And that's a distribution question, not a technology question.
• 20:04
  SPEAKER_00: And it gets answered by people in the rooms, not by a model.
• 20:07
  SPEAKER_00: I think it makes more of us entrepreneurs than employees because the cost of starting something just fell through the floor and it's not coming back up.
• 20:15
  SPEAKER_00: Some of us are impressed easily, and some of us are not impressed at all.
• 20:19
  SPEAKER_00: And every single one of us has to be a human being, the person standing in front of us, to the person standing in front of us.
• 20:27
  SPEAKER_00: 3,462 out of 52,881.
• 20:31
  SPEAKER_00: The sky is doing something.
• 20:32
  SPEAKER_00: It's not falling on the schedule you're being sold.
• 20:36
  SPEAKER_00: AI for everyone, not just the wealthy.
• 20:37
  SPEAKER_00: I'm Connor with honor.
• 20:39
  SPEAKER_00: Be safe out there and please have a good weekend.