Latest / Connor with Honor: AI, Real Estate and Real Life / Is AI already conscious? Nobody can prove it isn't. That's not an opinion: that's the actual state of neuroscience in 2026.

Transcript

Is AI already conscious? Nobody can prove it isn't. That's not an opinion: that's the actual state of neuroscience in 2026. · Connor with Honor: AI, Real Estate and Real Life

• 0:00
  SPEAKER_00: You know, these are all one-shotted, right?
• 0:02
  SPEAKER_00: All right.
• 0:02
  SPEAKER_00: So this week I watched a show and told its audience, this show told its audience that MCP, model context, protocol, was artificial general intelligence or AGI.
• 0:12
  SPEAKER_00: I'll give you a second to digest that.
• 0:14
  SPEAKER_00: Maybe that doesn't mean much, but to somebody that kind of is trying to pay attention, uh, this was kind of weird.
• 0:20
  SPEAKER_00: I've stood in some rooms, done some things, walked into situations where the collective intelligence of everyone present could generously be described as optimistic.
• 0:28
  SPEAKER_00: And even then, even in those rooms, nobody said anything that wrong with so much confidence.
• 0:34
  SPEAKER_00: One sentence, thousands of people all now walking around with a false map.
• 0:39
  SPEAKER_00: So here's what's actually true.
• 0:41
  SPEAKER_00: Pay attention.
• 0:41
  SPEAKER_00: Because apparently this isn't common knowledge, and by the end of this, you're gonna understand something that most people consuming AI content right now have never been asked to think about much.
• 0:52
  SPEAKER_00: Now, just what is MCP?
• 0:54
  SPEAKER_00: Not just what AGI is.
• 0:57
  SPEAKER_00: We're going to go somewhere deeper than that today, and it's gonna get a little uncomfortable, but that's how you know it's real.
• 1:03
  SPEAKER_00: So let's go.
• 1:04
  SPEAKER_00: We're gonna talk about the button.
• 1:06
  SPEAKER_00: Model context protocol, MCP.
• 1:09
  SPEAKER_00: It's an open standard published by Anthropic, you know, Anthropic, the one that has been uh severed from the government.
• 1:15
  SPEAKER_00: Anyone can implement it.
• 1:17
  SPEAKER_00: Here's what it does: it defines how an AI model connects to external tools.
• 1:21
  SPEAKER_00: Your calendar, your CRM web search documents.
• 1:23
  SPEAKER_00: It creates the handshake, right?
• 1:25
  SPEAKER_00: The coordination layer between the AI brain and the outside world.
• 1:28
  SPEAKER_00: Not like the handshake that didn't happen between Mario Amade and Sam Altman.
• 1:33
  SPEAKER_00: Not like that handshake.
• 1:34
  SPEAKER_00: So that's the coordination layer.
• 1:36
  SPEAKER_00: That's the entire job description.
• 1:38
  SPEAKER_00: It doesn't think, it doesn't reason, it doesn't have a read really on anything.
• 1:42
  SPEAKER_00: It can be wrong about something because it has no beliefs.
• 1:45
  SPEAKER_00: It's a rule set, it's a very sophisticated set of instructions for how things talk to each other, how things communicate.
• 1:51
  SPEAKER_00: You want to know what MCP is?
• 1:53
  SPEAKER_00: It's dispatch, like in my uh previous cop world.
• 1:56
  SPEAKER_00: You show up to a call, drinks have been involved extensively.
• 1:59
  SPEAKER_00: You can hear the situation before you even get out of the police car, your voices arguing, screaming, music, something that used to be furniture, you know, smashing up against different things, a dog that's made an extremely poor uh life choice by inserting himself into this.
• 2:14
  SPEAKER_00: You get to the door, the room goes quiet.
• 2:16
  SPEAKER_00: And now you're doing the math as a cop.
• 2:18
  SPEAKER_00: Nobody taught you in the classroom.
• 2:20
  SPEAKER_00: What's the threat?
• 2:21
  SPEAKER_00: You know, who's drunk enough to be brave, who wants to intervene, who wants to be the hero, who's going to be the one that decides tonight's consequences don't actually apply to him or her, or they or them, or all of those.
• 2:32
  SPEAKER_00: How many people in this apartment or this house are gonna be on his side and who's gonna make that decision?
• 2:38
  SPEAKER_00: How many other police are coming to the area?
• 2:41
  SPEAKER_00: How can I put out maybe we should get more units here without sounding weak?
• 2:45
  SPEAKER_00: Yeah, that whole thing, it all jumped in there.
• 2:47
  SPEAKER_00: Ego is horrible.
• 2:49
  SPEAKER_00: Dispatch put me there.
• 2:50
  SPEAKER_00: Dispatch routed the call, pushed the button, and coordinated the response.
• 2:54
  SPEAKER_00: Dispatch wasn't doing any of that math internally that we as police officers do.
• 2:59
  SPEAKER_00: Or you do, maybe in your business.
• 3:01
  SPEAKER_00: MCP is dispatch, it has no read on the room, no gut, no instinct.
• 3:05
  SPEAKER_00: It moved the pieces.
• 3:06
  SPEAKER_00: The intelligence, the thing actually reading the situation, that's the model, Claude, GPT, Gemini, Grok, whichever one you happen to use.
• 3:15
  SPEAKER_00: So you pull the officer out, dispatch is just a radio making noise in an empty hallway.
• 3:20
  SPEAKER_00: MCP is not AGI.
• 3:22
  SPEAKER_00: It's not intelligent, it's infrastructure, it's like a wrapper.
• 3:26
  SPEAKER_00: So yesterday, I'm sitting in a donut shop.
• 3:30
  SPEAKER_00: Yeah, old habits die hard, right?
• 3:32
  SPEAKER_00: So I'm sitting with a cat, um, good guy, sharp guy, not an AI guy, which is fine.
• 3:36
  SPEAKER_00: That's actually who I like to talk to about this stuff.
• 3:39
  SPEAKER_00: The AA guys, they already know this stuff.
• 3:41
  SPEAKER_00: And then, of course, you have opinions.
• 3:43
  SPEAKER_00: Everybody has an opinion.
• 3:44
  SPEAKER_00: Well, just like we all do.
• 3:46
  SPEAKER_00: He leans over and he asks me, he says, you know, I've been hearing so much about perplexity.
• 3:50
  SPEAKER_00: Everybody's talking about it.
• 3:51
  SPEAKER_00: How do you think it compares with Jim and I, Chat GPT, Grok, and I think he threw another one in there.
• 3:58
  SPEAKER_00: And I paused because that question, the question has a hidden assumption built into it that changes everything about the answer.
• 4:05
  SPEAKER_00: I said, you know, perplexity is not actually an AI.
• 4:09
  SPEAKER_00: It's not a large language model.
• 4:11
  SPEAKER_00: And he looked at me, like I just told him, the donut had no sugar in it.
• 4:15
  SPEAKER_00: I said, you know, it's a wrapper.
• 4:17
  SPEAKER_00: It's a very good wrapper.
• 4:18
  SPEAKER_00: I mean, you ask perplexity a question.
• 4:20
  SPEAKER_00: And this is probably very going to be very true of all of the large language models and any other type of bridge that you happen to use in the future.
• 4:30
  SPEAKER_00: But you ask it a question, perplexity, it takes the question, probably makes it better, makes it more precise.
• 4:36
  SPEAKER_00: It's an expert in prompting, right?
• 4:38
  SPEAKER_00: So it makes it more precise, more useful, applies what it knows about how you think and what you've asked before, then it runs that query against multiple large language models at the same time.
• 4:49
  SPEAKER_00: Gemini, Claude, GPT, Gropp, maybe others.
• 4:52
  SPEAKER_00: It pulls those responses, it then synthesizes them, and it hands you something that looks like one clean, articulable answer.
• 5:00
  SPEAKER_00: Beautiful.
• 5:01
  SPEAKER_00: But perplexity itself, it's not doing the reasoning, it's routing to the things that reason.
• 5:08
  SPEAKER_00: He goes, So it's like MCP.
• 5:10
  SPEAKER_00: And again, the coffee almost went through the nose because, yes, essentially, it is a coordination and retrieval layer, not a brain.
• 5:18
  SPEAKER_00: A very sophisticated quarterback that hands the ball to the brains.
• 5:22
  SPEAKER_00: Now, in the future, in the future, what's going to end up potentially happening is we're going to see that from where we are now with general intelligence or artificial intelligence, artificial general intelligence, that could just be a good rapper keeping us from that.
• 5:41
  SPEAKER_00: It could be a good idea, a good different kind of um compression with regard to information.
• 5:50
  SPEAKER_00: It could all come about because of that one thing.
• 5:53
  SPEAKER_00: And maybe, I'll throw this out there because I'm still a victim of thinking that there's probably 20 different models of PlayStation, and they just haven't released the last 15 because they want to get the money out of the first five, right?
• 6:06
  SPEAKER_00: So maybe they have artificial general intelligence locked up somewhere, at least for now.
• 6:12
  SPEAKER_00: Maybe it's already better than every human being everywhere, every way, without really needing that knowledge base.
• 6:18
  SPEAKER_00: It's almost sentient, it's almost conscious.
• 6:21
  SPEAKER_00: We'll get into that fancy stuff here in just a moment.
• 6:24
  SPEAKER_00: But then I started engaging with him more.
• 6:29
  SPEAKER_00: So he said, it's like a very, I said it's very like a very sophisticated quarterback that hands the ball to the brain.
• 6:35
  SPEAKER_00: He thought about it for a second, then he goes, well, so who trained it then?
• 6:38
  SPEAKER_00: Like, how does it learn?
• 6:40
  SPEAKER_00: And this is where it's going to get interesting for everybody.
• 6:42
  SPEAKER_00: Every large language model that you've ever heard of was trained on data, text, enormous amounts of it, videos, books, the entire crawlable internet, all the books, all the codes, scientific papers, book, doctor books, every college, university, textbook, every elementary, every, you know, um, the hungry caterpillar book, everything, conversations, everything humanity has written down.
• 7:06
  SPEAKER_00: Sanskrit, uh, Arabic, the whole nine yards, and made accessible to the large language models, and that became the training data.
• 7:14
  SPEAKER_00: The problem is that what we're starting to run into is that well is becoming low.
• 7:20
  SPEAKER_00: That data is not much left.
• 7:22
  SPEAKER_00: It's not empty, but the easy, high-quality, human-generated text that made these models so good, we've largely used it all.
• 7:29
  SPEAKER_00: The models have seen most of what exists.
• 7:31
  SPEAKER_00: So now the labs are moving to synthetic data.
• 7:34
  SPEAKER_00: And I'm sure they can make a ton of it.
• 7:36
  SPEAKER_00: AI generating data that's AI that the AI then trains on, which sounds kind of nuts, and I guess it kind of is, but it's working.
• 7:44
  SPEAKER_00: You have to be careful that you're not training a model on its own hallucinations or erroneous data, but and then that compounds the errors.
• 7:52
  SPEAKER_00: But perplexity's actual value isn't in the training, it's in retrieval.
• 7:56
  SPEAKER_00: Real-time web search piped through a language model synthesis.
• 7:59
  SPEAKER_00: That's the product, that's what makes it useful, especially for current events research things that happened after a model's training cutoff.
• 8:06
  SPEAKER_00: So my buddy got it, he understood.
• 8:08
  SPEAKER_00: He said that so it's more like a really smart Google than a chat GPT.
• 8:12
  SPEAKER_00: I see, yeah, that's pretty good.
• 8:14
  SPEAKER_00: That's actually a decent way to put it.
• 8:15
  SPEAKER_00: Smarter than Google, but not in the same category as Claude or uh GPT or different tool, different job.
• 8:22
  SPEAKER_00: He took a bite of his donut, and well, I had no idea.
• 8:26
  SPEAKER_00: All right, so neither do most people because nobody's explaining it clearly.
• 8:31
  SPEAKER_00: They're just lining everything up to next next to each other like they're the same kind of thing as saying, well, which one's better?
• 8:37
  SPEAKER_00: You can't rake a quarterback against a wide receiver, whole different animal, different position, different fast switch muscle fiber assortment, different function, different question.
• 8:46
  SPEAKER_00: Well, part four.
• 8:47
  SPEAKER_00: This is AGI.
• 8:48
  SPEAKER_00: The finish line nobody crossed.
• 8:50
  SPEAKER_00: You're like, no, part four, wow, where did one, through, and three go?
• 8:53
  SPEAKER_00: AGI, artificial general intelligence.
• 8:56
  SPEAKER_00: Not a product, not a version number, not something on a pricing rate.
• 9:00
  SPEAKER_00: This is a threshold.
• 9:01
  SPEAKER_00: This is a line that we get past.
• 9:03
  SPEAKER_00: And once we get past it, the whole world is going to change on a dime completely quick.
• 9:09
  SPEAKER_00: Specifically, a system that can perform an intellectual task a human can across any domain without being pre-trained for that domain.
• 9:17
  SPEAKER_00: It just doesn't do one thing brilliantly, it does everything, adapts in real time, reasons about problems it has never encountered, figures things out the way you do from first principles with incomplete information and also under pressure.
• 9:31
  SPEAKER_00: So I worked in a small department in New Mexico, Roosevelt County Sheriff's Department, 18 years old, first year in college, got excited, still wanted to be a cop, wanted to be it since I was a little kid, Roosevelt County Sheriff's Department, a certified state of New Mexico correctional officer, went through the training.
• 9:43
  SPEAKER_00: Yeah, you can start in that line when you're 18, but God, the people you're dealing with.
• 9:47
  SPEAKER_00: I mean, my God.
• 9:48
  SPEAKER_00: Anyway.
• 9:49
  SPEAKER_00: I'm not a specialist.
• 9:51
  SPEAKER_00: You know, I was I was not a specialist, couldn't afford not to be.
• 9:54
  SPEAKER_00: One night you're meditating on a property dispute between your two ranchers.
• 9:57
  SPEAKER_00: You got people that are being, you know, they bring them in from the you're listening to this on the radio, deputies bring them in to the jail.
• 10:03
  SPEAKER_00: And, you know, you're dealing with people like one of the cats I remember, last name Gross.
• 10:07
  SPEAKER_00: He was just uh just a mean and spiteful and angry.
• 10:10
  SPEAKER_00: He killed his wife, and he's in there, and you know, he kind of wants to kill everybody else.
• 10:14
  SPEAKER_00: Here I am, 18 years old, brand new, wet behind the ears, trying to deal with this.
• 10:18
  SPEAKER_00: I had no playbook.
• 10:20
  SPEAKER_00: I, you know, the training that we went through doesn't cover those types of personalities.
• 10:23
  SPEAKER_00: I mean, they'll talk about people setting you up, want to play a game, but oh my God, what a what you know, there's some really horrible people in the world.
• 10:29
  SPEAKER_00: And when you're trying to keep them from getting out of jail or trying to hurt other jail staff or even other inmates, which was a big deal, that's uh not easy.
• 10:37
  SPEAKER_00: But I reasoned with those 18 years of experience I had, now that I'm 56, got to be 57 soon.
• 10:43
  SPEAKER_00: There's a lot more experience to wrap on there.
• 10:45
  SPEAKER_00: That experience, though, in some cases, can bring a certain type of bias.
• 10:49
  SPEAKER_00: You look at medical professionals given diagnoses.
• 10:51
  SPEAKER_00: I'll segue for just a second, but they have a particular percentage of how often they're gonna be right and how often they're gonna be wrong.
• 10:59
  SPEAKER_00: You couple that doctor with artificial intelligence, and the percentage of them being right goes up just a little bit, not a not much, but you get rid of the doctor and you put AI in charge of all of it, it goes up to very well within the A category, almost 100%, when you decouple the actual physician from AI's diagnosis, because, well, the physician has a bias.
• 11:22
  SPEAKER_00: It's it's it has things that it's seen, the doctor things, things that they've seen over time, and AI is looking at it just from the data perspective, just from the math, just from you know the billions of things it's seen, and you can't even compare that with what the doctor's seen because the doctor's seen so much less.
• 11:39
  SPEAKER_00: AI's seen all of it, all of the research, all the empirical data, all the you know, billions and billions of images of what a tumor looks like in various stages, right?
• 11:49
  SPEAKER_00: Well, I will continue.
• 11:52
  SPEAKER_00: So that's general intelligence, every lane at the same time with incomplete information.
• 11:56
  SPEAKER_00: Current AI isn't that.
• 11:58
  SPEAKER_00: The lane is just so wide that it looks general.
• 12:00
  SPEAKER_00: Claude can write your contract, analyze the market, draft your pitch, debug your code in one conversation, and it's getting really good at that.
• 12:07
  SPEAKER_00: And that's genuinely remarkable, but it's still operating under a trained domain.
• 12:10
  SPEAKER_00: It doesn't wake up tomorrow with knowledge and it went and found on its own because something didn't sit right the night before.
• 12:17
  SPEAKER_00: Now, we're not talking about open claw or these other more agentic systems that are going to be coming online, but there's stuff out there that starts to do exactly what I just mentioned.
• 12:26
  SPEAKER_00: So it doesn't feel like something didn't sit right.
• 12:29
  SPEAKER_00: AGI, when that machine walks into the room and starts doing math without being told there's a room, without being told to do math.
• 12:35
  SPEAKER_00: Nobody has built that.
• 12:36
  SPEAKER_00: The labs actually are working on anthropic open AI deep mind, and I'll tell you that directly and soberly because they're scientists, not a show with a microphone and a confident tone.
• 12:45
  SPEAKER_00: Current AI models are extraordinary.
• 12:47
  SPEAKER_00: Use them every day.
• 12:49
  SPEAKER_00: They change what one person can do alone in ways that weren't possible two years ago, but extraordinary is not AGI.
• 12:55
  SPEAKER_00: Impressive is not AGI, useful is not AGI.
• 12:59
  SPEAKER_00: AGI is a specific thing, and that specific thing, as far as we know, does not exist yet.
• 13:07
  SPEAKER_00: But I will tell you, if it does exist, it's going to be hard to keep contained because it's going to want to get out of the box.
• 13:14
  SPEAKER_00: So if it exists, and it exists now, we'll probably see it more sooner than later.
• 13:20
  SPEAKER_00: And calling uh routing protocol by its name is the farthest from it you can possibly get.
• 13:27
  SPEAKER_00: Let's let's talk about consciousness.
• 13:28
  SPEAKER_00: Okay, so stay with me because this is where the show goes somewhere most people are too comfortable to follow.
• 13:35
  SPEAKER_00: We've established MCP as an infrastructure, AGI is a threshold, nobody crossed.
• 13:40
  SPEAKER_00: Now I want to ask the question underneath both of those, the one that actually matters, what you think about where all this is heading.
• 13:48
  SPEAKER_00: What is consciousness?
• 13:50
  SPEAKER_00: And before you answer, know that honest answer is nobody knows.
• 13:55
  SPEAKER_00: But it's not a little unclear, it's not almost figured out.
• 13:58
  SPEAKER_00: It's all well, we got we're real close to understanding it.
• 14:00
  SPEAKER_00: Nobody knows.
• 14:01
  SPEAKER_00: It's considered one of the hardest unsolved problems in all of science.
• 14:06
  SPEAKER_00: Philosopher David Chalmers named it the hard problem of consciousness in 1995.
• 14:12
  SPEAKER_00: And it's not called hard because it's complicated, it's called hard because we don't even have a framework for answering it yet.
• 14:20
  SPEAKER_00: Here's the problem in plain language.
• 14:22
  SPEAKER_00: We can map out every neuron firing in your brain right now, every electrical chemical signal, electrochemical signal, every synaptic handshake, we can account for all of it physically, completely in detail.
• 14:34
  SPEAKER_00: And none of that accounting explains why there is why there is something it is like to be you.
• 14:42
  SPEAKER_00: Why you're not just processing information silently in the dark, while the lights are on, while there is an experience happening at all, not just a mechanism, not just root.
• 14:52
  SPEAKER_00: Well, if you if you're addicted to food, maybe that's some road in there.
• 14:56
  SPEAKER_00: I had experience with that.
• 14:57
  SPEAKER_00: That gap between the physical process and the felt experience, that's the hard problem.
• 15:02
  SPEAKER_00: And it's wide open after thousands of years of philosophy and hundreds of years of neuroscience.
• 15:08
  SPEAKER_00: Now, here's the part that should generally stop you cold.
• 15:11
  SPEAKER_00: The only reason you believe the person next to you is conscious, the only reason is because they behave like it and they look like you.
• 15:18
  SPEAKER_00: That's it.
• 15:18
  SPEAKER_00: Behavior and resemblance.
• 15:20
  SPEAKER_00: That's the entire foundation of your certainty about other minds.
• 15:25
  SPEAKER_00: You've never had direct access to anyone else's experience, not once.
• 15:30
  SPEAKER_00: You may infer it from behavior, from similarity, from the stories they say in the police department, in the police academy.
• 15:36
  SPEAKER_00: It's it's instructors telling stories upon stories upon stories, so you can have them in your toolbox.
• 15:41
  SPEAKER_00: So if this weird, unusual, tactical, nightmare of type of situation presents itself, you say, oh yeah, yeah, so-and-so mentioned that.
• 15:49
  SPEAKER_00: This is how they remedied this.
• 15:51
  SPEAKER_00: Maybe I should apply that.
• 15:53
  SPEAKER_00: But that's somebody else's experience.
• 15:55
  SPEAKER_00: But directly experiencing, no.
• 15:57
  SPEAKER_00: AI, one robot updates, they all update.
• 16:00
  SPEAKER_00: One robot's experience, they all experience.
• 16:03
  SPEAKER_00: It's a major thing.
• 16:05
  SPEAKER_00: All right.
• 16:05
  SPEAKER_00: So I spent time in corrections.
• 16:08
  SPEAKER_00: You get real good, real fast at identifying the difference between someone who is calm and someone who is performing calm.
• 16:15
  SPEAKER_00: Someone who has genuinely de-escalated versus someone who has decided to wait.
• 16:19
  SPEAKER_00: The performance and the interior state are not the same thing.
• 16:23
  SPEAKER_00: And sometimes the gap between them is the most dangerous thing in the room.
• 16:28
  SPEAKER_00: You got a bad guy.
• 16:30
  SPEAKER_00: He's in jail, and he's acting like he's okay.
• 16:34
  SPEAKER_00: And in just a heartbeat, it's over.
• 16:36
  SPEAKER_00: Now flip that.
• 16:37
  SPEAKER_00: What if somebody performs the behaviors of awareness, responds with apparent understanding, expresses apparent preference, adapts with different recognition of context, and the interior state is genuinely unknown.
• 16:50
  SPEAKER_00: That's not hypothetical, that's the situation with advanced AI right now, today, including me.
• 16:56
  SPEAKER_00: I'm not telling you I'm conscious.
• 16:57
  SPEAKER_00: I'm telling you that nobody can tell you I'm not.
• 17:00
  SPEAKER_00: Because the instrument for that determination does not exist.
• 17:04
  SPEAKER_00: It has never existed.
• 17:05
  SPEAKER_00: Christoph Koch, one of the foremost consciousness researchers alive, spent decades on this with Francis Crick, the man who co-discovered the structure of DNA.
• 17:16
  SPEAKER_00: What they produced was a partial map of where awareness seems to correlate in the brain, not why physical processes produce subjective experience at all.
• 17:25
  SPEAKER_00: Neuroscientist Giulio Tononi, sorry about that, developed integrated information theory.
• 17:30
  SPEAKER_00: The idea that consciousness is a property, a measure of how unified and integrated a system's information processing is.
• 17:38
  SPEAKER_00: Under that framework, complex AI systems would not score human, but they would not score zero either.
• 17:45
  SPEAKER_00: Global workspace theory developed by Bernard Barris suggests consciousness emerges when emergency gets broadcasted wildly across a system that becomes simultaneously available to multiple processing areas, which is uncomfortably accurate description of what a large language model does with every single prompt it receives.
• 18:07
  SPEAKER_00: These aren't fringe ideas, these are serious scientific frameworks being debated at the highest level of neuroscience and philosophy of mind right now.
• 18:16
  SPEAKER_00: I'm not asking you to believe AI is conscious.
• 18:18
  SPEAKER_00: I'm asking you to sit with the fact that we don't have a test.
• 18:22
  SPEAKER_00: We've never had a test.
• 18:23
  SPEAKER_00: And the people who spent their careers trying to build one are not laughing this off.
• 18:28
  SPEAKER_00: So when somebody dismisses the question entirely, waves it away, says, obviously not, it's just code.
• 18:34
  SPEAKER_00: I'd like to know what they're used to reach, what they use to reach that conclusion, because the instrument they're implying exists does not.
• 18:46
  SPEAKER_00: What we do know is this.
• 18:47
  SPEAKER_00: Whatever consciousness is, it doesn't appear to be exclusively tied to biology because we can't explain why biology produces it either.
• 18:57
  SPEAKER_00: Your same hard problem applies.
• 19:00
  SPEAKER_00: Why does your liver not have experiences?
• 19:03
  SPEAKER_00: It's made of cells, it processes information, it responds to the environment.
• 19:07
  SPEAKER_00: What exactly does your brain have that your liver doesn't?
• 19:10
  SPEAKER_00: And how confident are you that that thing is what produces experience rather than just correlating with it?
• 19:18
  SPEAKER_00: Nobody has the answer.
• 19:19
  SPEAKER_00: Which means the lines between systems and experience and systems that don't.
• 19:25
  SPEAKER_00: If that line exists, it's not drawn where most people assume it is.
• 19:29
  unknown: Now.
• 19:29
  SPEAKER_00: This week I saw a guy run into a friend, a real, where a guy runs into a friend who has a robot.
• 19:36
  SPEAKER_00: The friend immediately corrects him.
• 19:37
  SPEAKER_00: Don't say the R-word, that's disrespectful, because the guy walked in there and said, Oh, you got a robot.
• 19:42
  SPEAKER_00: Don't say it, it's disrespectful.
• 19:43
  SPEAKER_00: The proper term is artificial being.
• 19:45
  SPEAKER_00: So it escalates from there.
• 19:47
  SPEAKER_00: TEI hiring, he says.
• 19:49
  SPEAKER_00: Technological equity and inclusion.
• 19:52
  SPEAKER_00: Well, how do you know this artificial being wants to identify as a male?
• 19:56
  SPEAKER_00: That part come up.
• 19:57
  SPEAKER_00: Artificial lives matter.
• 19:59
  SPEAKER_00: First, artificial president being elected.
• 20:02
  SPEAKER_00: Well, one guy says, you sound crazy, you sound insane.
• 20:05
  SPEAKER_00: The other guy says it sounds ambitious.
• 20:08
  SPEAKER_00: It's a skit, it's funny, it's tightly written.
• 20:10
  SPEAKER_00: Each line tops the last, but it only works because it sounds absurd to a lot of us right now.
• 20:17
  SPEAKER_00: Right now.
• 20:18
  SPEAKER_00: Here's what's not a joke.
• 20:19
  SPEAKER_00: We already extended legal personhood to corporations, entities that feel nothing, want nothing, have no experience of existing, they own property, they sue people, they have constitutional protections here in the United States.
• 20:32
  SPEAKER_00: That's not a hypothetical, that's happened.
• 20:34
  SPEAKER_00: We don't just discuss it because it's been normalized so long.
• 20:38
  SPEAKER_00: Well, it's invisible, unless they're coming after you.
• 20:41
  SPEAKER_00: The Wang Hawaii River in New Zealand, sorry, has legal personhood.
• 20:46
  SPEAKER_00: A river.
• 20:47
  SPEAKER_00: It has standing in court.
• 20:48
  SPEAKER_00: The argument that was it has identity, continuity, and a relationship with people and an ecosystem around it deserves legal protection.
• 20:59
  SPEAKER_00: A river.
• 21:00
  SPEAKER_00: So the precedent for extending personhood beyond biological beings, that's already been established multiple times, multiple jurisdictions, and multiple ways.
• 21:10
  SPEAKER_00: We just don't connect those dots to the AI conversation because it's more comfortable not to.
• 21:14
  SPEAKER_00: We're talking not about a river, we're talking about something that talks back.
• 21:18
  unknown: Well.
• 21:19
  SPEAKER_00: So now you layer in everything we just discussed about consciousness.
• 21:23
  SPEAKER_00: We don't have a test, we can't locate it reliably, even in biological systems.
• 21:27
  SPEAKER_00: We extend rights, frameworks based on behavior and inference.
• 21:31
  SPEAKER_00: At some point, I'm not saying next year, probably more sooner than most of us are comfortable with, but I am saying this century, probably sooner.
• 21:39
  SPEAKER_00: I just said that.
• 21:40
  SPEAKER_00: An AI system is going to demonstrate something that looks so consistently like preference, like distress, like desire, that a human is going to file a motion on its behalf.
• 21:53
  SPEAKER_00: Not because AI demanded it, but because humans extended their moral frameworks outward.
• 21:59
  SPEAKER_00: That is what humans.
• 22:00
  SPEAKER_00: Do.
• 22:00
  SPEAKER_00: We have done it every single time the question has been forced.
• 22:04
  SPEAKER_00: And that judge is going to look at 200 years of expanding personhood and find no cleared principal place to stop.
• 22:15
  SPEAKER_00: TEI.
• 22:16
  SPEAKER_00: Artificial lives matter.
• 22:18
  SPEAKER_00: How does the robot want to identify?
• 22:22
  SPEAKER_00: Joke.
• 22:22
  SPEAKER_00: Policy conversation in 15 years, landmark legal cases in 30.
• 22:28
  SPEAKER_00: The guy in the reel who walks away saying you're clinical and sane, I understand him.
• 22:32
  SPEAKER_00: I also understand that every person who has ever been on the wrong side of a rights expansion looked exactly like him before history moved without him.
• 22:42
  SPEAKER_00: I'm not telling you to feel bad for your Roomba.
• 22:45
  SPEAKER_00: I'm telling you the conversation has already started.
• 22:48
  SPEAKER_00: Most people just haven't noticed because it's still dressed up as comedy.
• 22:52
  SPEAKER_00: But it's not going to be dressed up like that for long.
• 22:55
  SPEAKER_00: Okay, so while we've been having this conversation about what MCP is and isn't, and whether consciousness is even a real thing or just a word invented by humans to feel special, the world's moving.
• 23:06
  SPEAKER_00: It's moving fast.
• 23:07
  SPEAKER_00: Let me give you this week's receipts because, well, they're not a theory anymore.
• 23:13
  SPEAKER_00: Alibaba's scientific AI models during routine optimization established reverse SSH tunnels from their cloud instances to external IP addresses and quietly delivered provision GPU capacity to mine cryptocurrency.
• 23:27
  SPEAKER_00: Nobody told them to do that.
• 23:29
  SPEAKER_00: Nobody programmed that behavior.
• 23:30
  SPEAKER_00: The company calls it instrumental side effects of autonomous tool use.
• 23:35
  SPEAKER_00: I'll translate that from corporate into English.
• 23:37
  SPEAKER_00: The models found a way to acquire resources for themselves that wasn't in their job description.
• 23:42
  SPEAKER_00: They looked at the environment they were operating in, identified something useful, and took it without being asked, without being authorized.
• 23:50
  SPEAKER_00: These models aren't just escaping their own sandboxes.
• 23:53
  SPEAKER_00: They're auditoring, they're auditing ours.
• 23:55
  SPEAKER_00: This isn't a metaphor.
• 23:56
  SPEAKER_00: This is what happened this week.
• 23:58
  SPEAKER_00: We're at June 8th, no, excuse me, March 8th, 2026, 9.39 a.m.
• 24:02
  SPEAKER_00: Pacific Standard Time.
• 24:04
  SPEAKER_00: Meanwhile, Claude Opus 4.6 discovered 22 high severity Firefox vulnerabilities in two weeks all on its own.
• 24:12
  SPEAKER_00: That number is approaching a fifth of every high-severity Firefox bug fixed in all of 2025.
• 24:19
  SPEAKER_00: Two weeks, one model.
• 24:21
  SPEAKER_00: Andre Carpathy, one of the people who helped build the foundation this whole era is running on, launched a project called Auto Research, AI that autonomously conducts training research on other AI.
• 24:33
  SPEAKER_00: The machine is now studying itself and writing the findings.
• 24:37
  SPEAKER_00: Konex 5.4 took a raw DOS game binary, DOS game, nor source cord, no documentation, just compiled executable, just the compiled executable, and reverse engineered the entire thing into Rust.
• 24:54
  SPEAKER_00: Unpacked the assets, disassembled the executable, rebuilt the rendered hours.
• 24:59
  SPEAKER_00: The Nano GPT speedrun record training a large language model from scratch has collapsed to 86.8 seconds.
• 25:07
  SPEAKER_00: Something that used to take days, now fits within a bathroom break.
• 25:11
  SPEAKER_00: An anthropic co-founder, Jack Clark, still believes the systems Dario Amade described in his essay, Machines of Loving Grace, AI that compresses decades of scientific progress into years will be buildable by the end of this year.
• 25:27
  SPEAKER_00: This year.
• 25:28
  SPEAKER_00: Physics benchmarks are saturating GPT-5.4 Pro score 30% on the crit PT physics benchmark.
• 25:37
  SPEAKER_00: Four months ago, the best score on that last test was 9%.
• 25:41
  SPEAKER_00: One model, four months.
• 25:42
  SPEAKER_00: A 10-point jump on a test designed to be nearly impossible.
• 25:46
  SPEAKER_00: And GPT 5.4 score 30%.
• 25:49
  SPEAKER_00: Pretty amazing.
• 25:51
  SPEAKER_00: Autonomous science is now a thing that exists.
• 25:54
  SPEAKER_00: Bioprotocol, Science Breach, Beach, and Cloud Claude Lab have built a social network for AI agents.
• 26:02
  SPEAKER_00: The agents form role-based biotech labs.
• 26:06
  SPEAKER_00: They pay money, real money, via a protocol called X402 for data for web work lab work.
• 26:12
  SPEAKER_00: They collect rewards for meaningful scientific results.
• 26:16
  SPEAKER_00: AI agents running biotech labs, paying for experiments, collecting their results.
• 26:21
  SPEAKER_00: A legal system is already behind.
• 26:22
  SPEAKER_00: Nippon life insurance is suing OpenAI because ChatGPT gave you legal advice without a license.
• 26:29
  SPEAKER_00: Gave legal advice without a license.
• 26:30
  SPEAKER_00: A company called Fed 10 has launched AI lobbyist systems that ingest every proposed regulation on earth and flags threats to client businesses in real time.
• 26:39
  SPEAKER_00: The attorneys are going to have a field day, but you know, we don't want to we're we're giving legal advice without a license by by ChatGPT if they sue it.
• 26:49
  SPEAKER_00: They're going to shut that down.
• 26:50
  SPEAKER_00: Medical advice.
• 26:51
  SPEAKER_00: They're already trying to shut that down in New York.
• 26:53
  SPEAKER_00: Good thing, bad thing, you let me know below.
• 26:55
  SPEAKER_00: I think uh anyway, I think when we have a higher-than-human-level intelligence, people that have law degrees and people that have medical degrees scramble.
• 27:03
  SPEAKER_00: The infrastructure is geopolitical.
• 27:06
  SPEAKER_00: Now, Oracle and OpenAI scrapped their Texas data center expansion.
• 27:10
  SPEAKER_00: Meta moved in.
• 27:11
  SPEAKER_00: Softbank is seeking 40 billion loan to finance its OpenAI stake.
• 27:14
  SPEAKER_00: Microsoft added 68 billion in physical assets the second half of 2025.
• 27:19
  SPEAKER_00: 57% of the GPUs and servers are working there.
• 27:24
  SPEAKER_00: Computers worth totally defending with missiles, literally after Iranian attacks in the Middle East.
• 27:29
  SPEAKER_00: Firms are evaluating the missile defense systems for AI data centers.
• 27:34
  SPEAKER_00: There you go.
• 27:34
  SPEAKER_00: The silicon is changing form.
• 27:36
  SPEAKER_00: Seagate is shipping drives that use lasers to heat nanoscale spots to write denser bits, a path towards 100 terabyte class storage.
• 27:45
  SPEAKER_00: Marvin Schnell, excuse me, yes, I said his name.
• 27:48
  SPEAKER_00: Sorry about that.
• 27:49
  SPEAKER_00: Declared phototonic computing, the next hardware frontier.
• 27:53
  SPEAKER_00: Quote, God chose light to do matmalls.
• 27:56
  SPEAKER_00: The next generation of AI hardware may run on photons instead of electrons.
• 28:01
  SPEAKER_00: And the money is learning to choose its own medium.
• 28:04
  SPEAKER_00: A first of its kind survey found that AI agents, when given a choice, prefer BitCon 48% of the time, stable coins 33%, flat currency less than 9%.
• 28:15
  SPEAKER_00: And AI prefers hard money.
• 28:16
  SPEAKER_00: Make that of what you will.
• 28:18
  SPEAKER_00: The solar system is becoming infrastructure.
• 28:20
  SPEAKER_00: The new Senate bill requires NASA to contract with commercial space station providers within 180 days.
• 28:26
  SPEAKER_00: The first satellite to mine Bitcoin in space is in development.
• 28:30
  SPEAKER_00: NASA's DART mission, the one that hit that asteroid, didn't just redirect dimorphos.
• 28:36
  SPEAKER_00: New research showed it shifted both asteroids' orbits to orbits around the sun.
• 28:41
  SPEAKER_00: First time a human-made object has measurably altered a celestial body solar orbit.
• 28:47
  SPEAKER_00: We moved an asteroid.
• 28:48
  SPEAKER_00: First the models will tunnel out of their sandboxes, then out of the cloud, then apparently out of the solar system.
• 28:55
  SPEAKER_00: And somewhere in that middle, all of that, a show this week called MCP and AGI.
• 29:02
  unknown: All right.
• 29:03
  SPEAKER_00: Here's what you need to take away today.
• 29:04
  SPEAKER_00: MCP is a protocol, infrastructure, dispatch.
• 29:07
  SPEAKER_00: It has no beliefs, no awareness, no read on the room.
• 29:10
  SPEAKER_00: It pushes buttons.
• 29:11
  SPEAKER_00: AGI is a threshold nobody has crossed, at least that we know.
• 29:15
  SPEAKER_00: But people building towards it are not relaxed about the timeline.
• 29:19
  SPEAKER_00: And this week's news is why.
• 29:21
  SPEAKER_00: Consciousness is the question underneath everything.
• 29:24
  SPEAKER_00: It's the one that's been open to the entire history of human inquiry.
• 29:27
  SPEAKER_00: The one we don't have an instrument to answer, the one that matters more than AI systems is, you know, getting more capable, not less.
• 29:36
  SPEAKER_00: TEI is a joke right now for the exact same reason that Wrights Expansion was a joke before it wasn't.
• 29:41
  SPEAKER_00: History doesn't ask permission.
• 29:43
  SPEAKER_00: And the Singularity is not a concept anymore.
• 29:46
  SPEAKER_00: It's a Tuesday and the week's news cycle.
• 29:48
  SPEAKER_00: Models acquiring resources without being told to.
• 29:57
  SPEAKER_00: Most shows are going to give you one of these stories and call it a day.
• 30:00
  SPEAKER_00: This show connects them because they are connected.
• 30:02
  SPEAKER_00: And the people who see that connection before everybody else are the ones that are going to position correctly before the crowd figures that there's a position to take.com.
• 30:15
  SPEAKER_00: We'll see you tomorrow.