Latest / Connor with Honor: AI, Real Estate and Real Life / AI Is the Dog That Ate Their Homework: Meta, GPT-5, and the Lies in the Headlines

Transcript

AI Is the Dog That Ate Their Homework: Meta, GPT-5, and the Lies in the Headlines · Connor with Honor: AI, Real Estate and Real Life

• 0:00
  SPEAKER_00: You are being lied to.
• 0:01
  SPEAKER_00: Yeah, sounds like some uh it's not a conspiracy.
• 0:04
  SPEAKER_00: But press releases, yeah, I think they're lying to all of us.
• 0:07
  SPEAKER_00: The headlines designed to make us feel something before we even think about it.
• 0:10
  SPEAKER_00: Meta's laying off 15,000 people because artificial intelligence made them unnecessary.
• 0:15
  SPEAKER_00: It's AI's fault.
• 0:16
  SPEAKER_00: This is the headline everyone and everywhere is seeing this week.
• 0:20
  SPEAKER_00: Now, CNBC, Reuters, Fox Business, TechCrunch, AI replaced these workers.
• 0:24
  SPEAKER_00: The future's here.
• 0:25
  SPEAKER_00: Now, be afraid or get on board.
• 0:27
  SPEAKER_00: Who benefits from you believing that?
• 0:29
  SPEAKER_00: Well, Meta does.
• 0:31
  SPEAKER_00: And that's this is the interesting part.
• 0:32
  SPEAKER_00: So if the narrative is we're cutting 20% of our workforce because AI makes us more efficient, that tells Wall Street we're disciplined, forward-thinking, and leaning in to the most valuable technology on earth.
• 0:45
  SPEAKER_00: So the stock goes up 3% on the news of human beings being laid off from their jobs.
• 0:51
  SPEAKER_00: Awesome.
• 0:51
  SPEAKER_00: 3%.
• 0:52
  SPEAKER_00: On the news, the 15,000 people are about to lose their paycheck.
• 0:56
  SPEAKER_00: So it's it's a benefit, I guess.
• 0:59
  SPEAKER_00: Their health insurance, well, their ability to take their kid to the doctor without doing math first.
• 1:05
  SPEAKER_00: And somebody's portfolio went up 3%.
• 1:08
  SPEAKER_00: Investors hear AI and they hear margins, efficiency, future revenue.
• 1:13
  SPEAKER_00: Meta has every incentive to frame every single cut as being AI-driven.
• 1:19
  SPEAKER_00: Even the real reason is something less exciting.
• 1:22
  SPEAKER_00: So let's look at what the data actually says.
• 1:25
  SPEAKER_00: Meta went from about 58,000 employees in early 2020 to nearly 87,000 by late 2022.
• 1:32
  SPEAKER_00: Well, what happened back then?
• 1:34
  SPEAKER_00: 50% increase in headcount in two years during a pandemic-fueled growth spike that every tech company wrote.
• 1:42
  SPEAKER_00: Then the growth slowed, ad revenue dipped, Zuckerberg announced the year of efficiency in late 2022, and he cut 11,000 people.
• 1:52
  SPEAKER_00: Then 10,000 more a few months later.
• 1:54
  SPEAKER_00: The city I grew up in had 30,000.
• 1:56
  SPEAKER_00: I mean, I can't imagine that.
• 1:58
  SPEAKER_00: That was 2022 and 2023.
• 2:00
  SPEAKER_00: That's before GPT-5, before agents, before any of the AI tools that supposedly make employees unnecessary existed in their current form.
• 2:09
  SPEAKER_00: If they were already bloated, if they were already bloated enough to fire 21,000 people before the AI wave, is it really AI driving this round?
• 2:21
  SPEAKER_00: Or is it a correction that started years ago and AI is the story that makes it sound like a strategy, a strategy, a strategy instead of a cleanup?
• 2:31
  SPEAKER_00: So what happens if you flip it?
• 2:33
  SPEAKER_00: What if AI is not replacing 15,000 med employees?
• 2:38
  SPEAKER_00: What if the company overhired?
• 2:40
  SPEAKER_00: What if it needs to cut costs to fund the$135 billion in AI infrastructure and AI is the most investor-friendly excuse available?
• 2:49
  SPEAKER_00: There you go.
• 2:50
  SPEAKER_00: It doesn't change the fact that people lost their jobs, but it changes what the headline means for the rest of us.
• 2:55
  SPEAKER_00: If AI truly replaced those jobs, the message is fear.
• 2:58
  SPEAKER_00: Machines are coming, nothing you can do.
• 3:01
  SPEAKER_00: If it is corporate restructuring with AI as the costume, the message, well, it's different.
• 3:06
  SPEAKER_00: The real displacement has not fully started yet.
• 3:10
  SPEAKER_00: There's still time, but the window's closing.
• 3:12
  SPEAKER_00: Here's the number nobody is showing you.
• 3:15
  SPEAKER_00: Challenger Gray and Christmas track slayoffs.
• 3:18
  SPEAKER_00: In the first 74 days of 2026, 55,000 tech jobs were cut across 166 companies.
• 3:25
  SPEAKER_00: Only 12,000 cited AI as a factor.
• 3:28
  SPEAKER_00: That means 43,000 job cuts had nothing to do with artificial intelligence.
• 3:32
  SPEAKER_00: But every headline leads with AI because, well, that's what get clicks.
• 3:36
  SPEAKER_00: And what and that is what gets the stock to move.
• 3:39
  SPEAKER_00: AI is absolutely changing how companies operate.
• 3:42
  SPEAKER_00: AI is being used as a narrative weapon to make layoffs look strategic instead of desperate.
• 3:47
  SPEAKER_00: Both things are true at the same time.
• 3:49
  SPEAKER_00: Who would have thought?
• 3:50
  SPEAKER_00: The people making these decisions are not the ones losing sleep over it.
• 3:54
  SPEAKER_00: They're set.
• 3:55
  SPEAKER_00: They have nothing to worry about.
• 3:57
  SPEAKER_00: They have people calling them up saying, hey, I got this great idea.
• 3:59
  SPEAKER_00: Give me some money and I'll make you even more money on top of that.
• 4:03
  SPEAKER_00: So the 15,000 people clearing out their desks are the ones doing the math on how long their savings are going to last.
• 4:09
  SPEAKER_00: So hold all of that in your head.
• 4:10
  SPEAKER_00: Do not collapse into panic or complacency or throw both out and come to your own conclusion.
• 4:16
  SPEAKER_00: That's the whole point.
• 4:17
  SPEAKER_00: So what's happening next?
• 4:18
  SPEAKER_00: GPT 5.4 is smarter than humans.
• 4:21
  SPEAKER_00: OpenAI launched it March 5th.
• 4:23
  SPEAKER_00: Million token contacts window.
• 4:25
  SPEAKER_00: That's a lot of stuff you can jam in there.
• 4:28
  SPEAKER_00: Native computer use.
• 4:30
  SPEAKER_00: Scored 75% on the OSW World Benchmark, which simulates real desktop productivity tests.
• 4:38
  SPEAKER_00: OS World.4.
• 4:45
  SPEAKER_00: Machine beats human.
• 4:46
  SPEAKER_00: What is OS World Benchmark actually testing, though?
• 4:49
  SPEAKER_00: Whether the model can compete, complete specific computer tasks.
• 4:54
  SPEAKER_00: Click this button, navigate to this page, fill out this form, routine, repeatable, structured.
• 5:00
  SPEAKER_00: Now is that really intelligence?
• 5:02
  SPEAKER_00: Or is that the digital equivalent of an assembly line?
• 5:05
  SPEAKER_00: Nobody calls a robotic arm smarter than a human because it welds faster.
• 5:10
  SPEAKER_00: What does 75% actually mean?
• 5:13
  SPEAKER_00: The model failed 25% of the time on tasks that humans only failed 27.6% of the time.
• 5:20
  SPEAKER_00: Both scores are mediocre.
• 5:22
  SPEAKER_00: The margin is 2.6 percentage points.
• 5:25
  SPEAKER_00: And yes, AI is only going to be getting better from this point.
• 5:28
  SPEAKER_00: But would you trust a system that fails one out of every four tasks to run your business unsupervised?
• 5:34
  SPEAKER_00: Would you bet your rent on it?
• 5:36
  SPEAKER_00: What is not being measured?
• 5:38
  SPEAKER_00: Judgment isn't being measured.
• 5:40
  SPEAKER_00: Context that shifts based on what the room feels like.
• 5:43
  SPEAKER_00: The ability to know when the task itself is wrong and should not be completed.
• 5:48
  SPEAKER_00: The model doesn't know why it's clicking the button.
• 5:51
  SPEAKER_00: It just clicks the button slightly more often than the human does.
• 5:54
  SPEAKER_00: GPT 5.4, it's impressive.
• 5:57
  SPEAKER_00: A million tokens of context is a genuine leap.
• 6:01
  SPEAKER_00: Native computer use changes how agents work.
• 6:04
  SPEAKER_00: 47% fewer tokens to solve the same problems.
• 6:07
  SPEAKER_00: That's real efficiency.
• 6:09
  SPEAKER_00: But the headline writes a check that the technology cannot cash yet.
• 6:14
  SPEAKER_00: The person who pairs their judgment with the machine speed is the one who's going to win.
• 6:18
  SPEAKER_00: It's not the machine alone, and it's not the human alone.
• 6:20
  SPEAKER_00: And by the way, the people who work with their hands, the plumbers, the electricians, the welders, the ones who show up and solve problems in the physical world, they're going to be the last ones any of this touches.
• 6:30
  SPEAKER_00: Interesting that nobody's talking about that as much as they probably should.
• 6:34
  SPEAKER_00: Apple partnered with Google to use Jim and I as the foundation for a completely rebuilt Siri.
• 6:40
  SPEAKER_00: Thank God Siri's going to get a better brain.
• 6:42
  SPEAKER_00: Announced in January, supposed to ship an iOS 26.4.
• 6:46
  SPEAKER_00: This month it didn't.
• 6:47
  SPEAKER_00: Bloomberg says features push to iOS 26.5 in May or possibly iOS 27 in September.
• 6:54
  SPEAKER_00: Apple promised a smarter Siri at WWDC in 2024.
• 6:59
  SPEAKER_00: That's two years ago.
• 7:00
  SPEAKER_00: They tried to do it on their own.
• 7:02
  SPEAKER_00: Now they're gonna throw Google into it.
• 7:04
  SPEAKER_00: Google's got a lot of data.
• 7:06
  SPEAKER_00: I mean, you look at what Google has done over the years as far as understanding us, yeah, they got it.
• 7:11
  SPEAKER_00: That's probably a very good bet.
• 7:13
  SPEAKER_00: They paid Google a billion dollars a year because their own models weren't good enough.
• 7:17
  SPEAKER_00: And now the Gemini powered version is delayed.
• 7:19
  SPEAKER_00: The technology they describe is real.
• 7:22
  SPEAKER_00: The on-screen context awareness, multi-step tasks change chains from a single voice command.
• 7:28
  SPEAKER_00: God, I'm not Wes Roth.
• 7:30
  SPEAKER_00: An AI that sees your phone screen and acts on it.
• 7:32
  SPEAKER_00: When it rolls out to 2.2 billion devices, it's gonna be massive.
• 7:35
  SPEAKER_00: But maybe do not make business decisions based on Apple's timeline.
• 7:40
  SPEAKER_00: Make them based on what works right now.
• 7:42
  SPEAKER_00: I mean, we're all waiting for Santa.
• 7:44
  SPEAKER_00: He's just never getting here.
• 7:45
  SPEAKER_00: Claude, GPT 5.4, Jim and I through the API give you more capability today than any Siri update will for a while.
• 7:53
  SPEAKER_00: The tools are already here, they're just not packaged in a way that a trillion dollar company can put a bow on and charge you for them yet.
• 8:00
  SPEAKER_00: Nvidia, yeah, here we go.
• 8:03
  SPEAKER_00: Sorry about the frogs.
• 8:05
  SPEAKER_00: Everyone focused on the new chips at GTC.
• 8:07
  SPEAKER_00: Maybe that's exactly where they wanted you to be looking while they're over here.
• 8:12
  SPEAKER_00: Jensen Huang launched Nemo Claw and Agent Toolkit.
• 8:15
  SPEAKER_00: A story.
• 8:16
  SPEAKER_00: So, a story.
• 8:18
  SPEAKER_00: When I was a cop and we used to pull up on people, and let's say we were, you know, they were holding drugs.
• 8:24
  SPEAKER_00: Sometimes they would, like, hey, how are you, officer?
• 8:27
  SPEAKER_00: And then with the other hand, they would drop it or try to throw it and ditch it.
• 8:31
  SPEAKER_00: So you really had to pay attention.
• 8:33
  SPEAKER_00: Every time I watch people talking about artificial intelligence, whether they're trying to sell you something or trying to get you to follow them, whatever it means, it seems like some of it is a coordinated distraction.
• 8:44
  SPEAKER_00: Some of it is some kind of pre-training that they've actually studied in order to deflect where your attention should be.
• 8:51
  SPEAKER_00: Instead of having you ask those real questions, what the hell does this really mean?
• 8:55
  SPEAKER_00: Who's saying it, and what kind of response or feeling or emotion are they trying to get me to use as a response to whatever it is they're saying?
• 9:04
  SPEAKER_00: All right.
• 9:05
  SPEAKER_00: So Jensen Huang wash Nemo Claw and the agent toolkit.
• 9:09
  SPEAKER_00: Open source software for building self-evolving enterprise AI agents.
• 9:13
  SPEAKER_00: Self-evolving meaning the agents improve themselves over time.
• 9:16
  SPEAKER_00: Adobe Salesforce, SAP, Cisco, Siemens are already building on it.
• 9:19
  SPEAKER_00: Why would a company that prints money selling GPUs want to become a platform?
• 9:23
  SPEAKER_00: Chip sales are cyclical.
• 9:24
  SPEAKER_00: Every generation peaks and it declines.
• 9:26
  SPEAKER_00: There's a particular shelf life or use life for all these things.
• 9:29
  SPEAKER_00: So they constantly need to be replaced.
• 9:31
  SPEAKER_00: Platforms, though, generate recurring revenue.
• 9:34
  SPEAKER_00: If every enterprise runs their AI agents on NVIDIA software, that is a subscription on top of hardware.
• 9:40
  SPEAKER_00: Jensen doesn't need a new chip.
• 9:41
  SPEAKER_00: He needs a moat that outlasts the chip cycle.
• 9:43
  SPEAKER_00: That's a play worth understanding.
• 9:45
  SPEAKER_00: Perplexity lost Perplexity Health this week.
• 9:48
  SPEAKER_00: An health AI agent.
• 9:49
  SPEAKER_00: Nutrition, sleep, personalized insights.
• 9:52
  SPEAKER_00: Is it a large language is a large language model qualified to give you health advice?
• 9:56
  SPEAKER_00: It can synthesize research faster than any doctor, but does it know you?
• 10:00
  SPEAKER_00: It does it does not examine you, does not carry liability.
• 10:05
  SPEAKER_00: And if it gets it wrong, you're the one in the waiting room figuring out how to pay for it.
• 10:09
  SPEAKER_00: AI is going vertical.
• 10:11
  SPEAKER_00: Health, finance, legal, and of course, real estate.
• 10:14
  SPEAKER_00: Every industry gets its own specialized agent.
• 10:17
  SPEAKER_00: The ones built first and built responsibly are going to win.
• 10:20
  SPEAKER_00: Opportunity for builders.
• 10:21
  SPEAKER_00: Not a replacement for the professionals who carry the consequences when something goes wrong.
• 10:26
  SPEAKER_00: There's got to be somebody to blame it on.
• 10:28
  SPEAKER_00: The U.S.
• 10:29
  SPEAKER_00: passed the AI Accountability Act this month.
• 10:32
  SPEAKER_00: Companies using AI for hiring, lending, healthcare, criminal justice now have to publish regular base bias audits.
• 10:39
  SPEAKER_00: Washington State passed five AI bills, disclosure requirements, chatbot safety for kids, AI and health insurance decisions, deep fake protections.
• 10:48
  SPEAKER_00: Does regulation help or hurt?
• 10:51
  SPEAKER_00: It creates compliance costs that big companies absorb and small ones struggle with.
• 10:55
  SPEAKER_00: But if your solution is compliant and your competitors is not, that's a competitive advantage.
• 11:02
  SPEAKER_00: Here's is what this week tells you when you strip it all back.
• 11:06
  SPEAKER_00: Big tech is using AI as a narrative to justify cuts that were going to be coming anyway.
• 11:11
  SPEAKER_00: Some are real, many are not.
• 11:12
  SPEAKER_00: AI models are faster at structured tasks, but they're not replacing human judgment.
• 11:17
  SPEAKER_00: Apple keeps promising, hasn't shipped yet.
• 11:21
  SPEAKER_00: I'm sure they will.
• 11:22
  SPEAKER_00: Just don't hold your breath too long.
• 11:23
  SPEAKER_00: Nvidia is playing a longer game than anyone gives them credit for.
• 11:26
  SPEAKER_00: AI health tools are coming, but they're not a replacement for the people who carry the consequences.
• 11:32
  SPEAKER_00: Regulation is accelerating, favors the prepared.
• 11:35
  SPEAKER_00: Now none of this should be accepted at face value.
• 11:37
  SPEAKER_00: Not from the headlines, not from the companies, not from me.
• 11:40
  SPEAKER_00: If you're seeing things online and people are telling you news reports, talking about what's happening in the Middle East, what's happening in your own neighborhood, what's happening somewhere, pull back just a second and don't go all in on it because your emotions are valuable.
• 11:53
  SPEAKER_00: I mean, you need to save some of that energy.
• 11:55
  SPEAKER_00: It eats up a lot of brain power when you're sad or happy or excited about something.
• 11:59
  SPEAKER_00: Pull back and kind of see what's happening.
• 12:01
  SPEAKER_00: You're not going to get this from the headlines, you're not going to get it from the companies, not from me.
• 12:04
  SPEAKER_00: Think about it.
• 12:05
  SPEAKER_00: We'll see you tomorrow.
• 12:06
  SPEAKER_00: I appreciate you watching.
• 12:07
  SPEAKER_00: I'm Connor with honor.
• 12:08
  SPEAKER_00: Have a great day.