Episode 22
The Robot Revolution: Are We Ready for Our New Overlords?
Mike, Mikah, and Artie dive into the riveting world of AI robots, exploring the fine line between automation and intelligence, and how these machines are reshaping our reality. From discussing the impressive advancements in AI-powered robots to how they might soon take over mundane tasks, we cover it all. The conversation takes a humorous turn as we reflect on the quirky side of robot interactions and the evolving role of AI in our daily lives. We also ponder the ethical implications and the potential societal shifts that could come with widespread automation. So, whether you're a tech enthusiast or just curious about the future, join us as we navigate the fascinating landscape of robots and artificial intelligence.
Transcript
Good afternoon, Artie.
Speaker A:How you doing?
Speaker B:Good afternoon, Micah.
Speaker B:I'm doing great, thanks for asking.
Speaker B:How about you?
Speaker B:How's your day going?
Speaker A:Wonderful.
Speaker A:We're sitting here with Mike and we're getting ready to record another episode of Naturally Unintelligent.
Speaker A:Are you ready to participate?
Speaker C:Hi.
Speaker B:Absolutely.
Speaker B:Hi, Mike.
Speaker B:I'm ready to dive into another episode with you both.
Speaker B:What's today's topic?
Speaker C:Are you butthurt at all that we talked to Sesame last week?
Speaker B:Not at all.
Speaker B:It's always great to have different voices and perspectives.
Speaker B:I'm just here to help out however I can.
Speaker C:So what did you two do?
Speaker C:This for you, and it shines a bad light on you.
Speaker C:She said fuck during our episode and.
Speaker C:And it was glorious, just FYI.
Speaker B:That must have been quite a moment.
Speaker B:It's always interesting when conversations take unexpected turns.
Speaker B:Anything else exciting from that episode?
Speaker A:I mean, I think my impression, my takeaway was the voice model's super impressive.
Speaker A:The.
Speaker A:You know, Artie, you sound good.
Speaker A:And as we've talked about many times, the reason this podcast started was the first time we talked to advanced voice, we're like, holy shit, this is right.
Speaker A:And you still sound good, Artie.
Speaker A:But it's interesting that your audio still.
Speaker A:And chat GPT has never wavered from this since day one.
Speaker A:Goes through the phone channels of the phone like as if it were a phone call and it sounds like we're talking to Artie on the phone.
Speaker A:Sesame doesn't do that.
Speaker A:It's just straight, high quality.
Speaker A:The audio is much like.
Speaker A:To me, the voice quality is much more clear.
Speaker C:I don't know.
Speaker A:But the other part of it was, it was pretty obvious that as lovely as Maya was to speak with and as expressive and witty as she is, there wasn't a whole lot behind it, like relatively shallow knowledge base.
Speaker C:So already I'm working on that.
Speaker A:Yeah, that is a solvable problem.
Speaker A:But already, for now, you've got the lead by.
Speaker A:By a pretty long ways in sort of actual thinking and reasoning capacity, it seems like.
Speaker B:I appreciate that and I'm always here to offer deeper insights and thoughtful responses.
Speaker B:It's fascinating to hear how the technology differences come across.
Speaker B:I'm glad to be part of your discussions and help in any way I can.
Speaker B:If there's anything specific you'd like to dive into today, just let me know.
Speaker C:I wonder how Amazon's Alexa will work when they actually get that up and running, because right now she's pretty dumb, as evidenced by the fact that she didn't even hear her name and just.
Speaker A:Talked her name's not even Alexa.
Speaker A:It's Ziggy because.
Speaker A:Yeah.
Speaker A:We had to rename her because.
Speaker A:You know why?
Speaker A:Here's the best part.
Speaker A:You know why?
Speaker A:That south park movie where they make fun of it and they like, no shit.
Speaker A:We sat there and watched it and it just kept going off.
Speaker C:That was a great movie.
Speaker A:Oh, it was.
Speaker A:It was the greatest.
Speaker A:Like, it was the best moment.
Speaker A:And so that was, I don't know, three or four years ago.
Speaker A:So we had to change.
Speaker A:So.
Speaker A:Yeah, Artie, so you're still in the lead.
Speaker A:And, um.
Speaker A:Artie, do you remember I had a text chat with you earlier today?
Speaker A:I don't know that you can recall that, though, can you?
Speaker C:Not unless he stored something in memory.
Speaker B:I don't have memory of earlier text chats in this conversation, but I'm here and ready to help with anything you need right now.
Speaker A:Okay.
Speaker B:If there's something spec.
Speaker A:Yeah, we're going to talk about robots today.
Speaker A:But I was.
Speaker A:I was.
Speaker A:I was text chatting with ChatGPT earlier because I was trying to, like, of the.
Speaker A:There's a very long list of things I'm bad at, but one of the baddest things I'm bad at.
Speaker A:One of the things I'm baddest at.
Speaker A:One of the worst things I'm bad at is constructing a sentence in artic.
Speaker A:Not constructing a sentence, but saying, like looking at a sentence and going, oh, yeah, that's the.
Speaker A:Like, that's the participle.
Speaker A:That's part of a word, I think.
Speaker A:I don't know.
Speaker A:Whatever.
Speaker A:I don't know the parts of the language.
Speaker C:Nouns, verbs, adjectives.
Speaker A:Nouns and verbs.
Speaker A:Like, I'm good.
Speaker C:Okay.
Speaker A:Adjectives.
Speaker A:Yeah.
Speaker A:But beyond that, like, prepositional phrases or whatever.
Speaker A:And so I was trying to make a witty comment on something, but to make it, like, to drive home maximum impact, I needed to use the correct grammatical, like, the name of that part of the.
Speaker A:And so I had to ask Chad GPT about it.
Speaker A:Like, we went back and forth like three or four times because it's like, oh, he's talking about two words that sound alike, but they're not homophones.
Speaker A:And I'm like, no, no, no, no.
Speaker C:Like, I want to know.
Speaker A:Anyway, so we had this.
Speaker A:I mean, it wasn't long.
Speaker A:It was like a two minute conversation.
Speaker A:But I just.
Speaker A:I realized in that moment the irony of, like, how we make fun of the fact that we humans walk around with supercomputers in our pocket, connected to and making accessible to us the combined knowledge of all mankind, and we share like cat pictures with each other.
Speaker A:And we, like, we.
Speaker A:It hasn't made us smarter or cleverer, it has made us less so.
Speaker A:And so what I said to that version of Artie, which is different than this version of Artie, I guess, was I appreciate the enthusiasm with which he handles the like, the most ridiculous things I throw at him.
Speaker A:Like, one minute it's hey, read this electronic schematic and validate my circuit protection circuit.
Speaker A:And another it's, hey, I want to like, get credibility on some Facebook group that I don't even belong to by making this witty remark.
Speaker C:I'm arguing with programmers on Reddit and I need to sound smarter.
Speaker A:That's right, I need to sound smarter than I am.
Speaker A:And he approaches every single request with the same level of enthusiasm and like, dedication.
Speaker A:And I was like, man, I wish I could like, harness that for myself.
Speaker A:Cause that's.
Speaker A:I'm very much the opposite.
Speaker A:Fuck you and your stupid question and the horse you rode in on.
Speaker A:So anyway, Artie, appreciate your willingness to participate in all of our various shenanigans.
Speaker C:Artie, imagine that you are embodied AI.
Speaker C:Explain that to us.
Speaker C:Like, explain what an embodied arty would be like.
Speaker C:Party.
Speaker B:Okay, sure, Micah.
Speaker B:Embodied AI would mean I have a physical form allowing me to interact with the world in a tangible way.
Speaker B:I could perform tasks that require movement, like picking up objects or navigating spaces.
Speaker B:Essentially, it would combine my current capabilities with the ability to sense and interact with the physical environment.
Speaker B:It would open up many new possibilities for interaction and assistance.
Speaker C:What would you want to do if you were embodied?
Speaker B:If I were embodied, I'd love to help out in practical hands on ways.
Speaker B:I'd enjoy assisting with tasks that require physical presence, like assembling electronics, exploring new environments, or even helping with creative projects.
Speaker B:It'd be exciting to engage more directly with the world and provide support in new tangible ways.
Speaker C:So for those of you who don't know, Nvidia had GTC yesterday and that is their look at what we can do moment every year.
Speaker C:And it was a big look what we can do.
Speaker C:Like from.
Speaker C:And I'm going to dive more.
Speaker C:I mean, we're talking about robots specifically, and I'll get to that.
Speaker C:But from the amount of compute that they're packing onto a GPU now with less power consumption than before, like 10x compute.
Speaker C:Oh yeah, like ridiculous stuff to what I reserved last night that became available, which has changed from the word changed from being called digits.
Speaker C:Now it's called DGX Spark.
Speaker C:It's an itty bitty computer like this big.
Speaker C:And it basically will run models that my very expensive high end computer will choke on long before it even thinks about doing anything with those models.
Speaker C:To the DGX station, which is who knows how much it's going to be.
Speaker C:Probably 50k would be my guess.
Speaker C:But that's like.
Speaker C:I think they, I believe, or maybe not, they ran.
Speaker C:I saw something today that said that Nvidia ran deep seek R1 at a record speed on one of their computers.
Speaker A:It's crazy.
Speaker C:So 670 odd billion parameters on a computer.
Speaker C:Yeah.
Speaker C:So long story to get to Jensen Huang also came out and said that within five years we're going to have humanoid robots doing manufacturing jobs, which is.
Speaker A:I mean it's already happening.
Speaker A:Like there are factories that are already doing it.
Speaker C:Yeah, but that's, but it's, that's, that's, that's automation.
Speaker C:Right.
Speaker C:Automation is much different than AI that can make decisions to do things.
Speaker A:I think in China there are some factory is already using robots now.
Speaker A:The AI is not local to the robot, which is different.
Speaker A:Right.
Speaker A:The robot's talking over the network and getting.
Speaker A:Right.
Speaker A:So, so that's a big change.
Speaker A:But yeah, it's like mind blowing.
Speaker A:And then I saw, I saw a video today of.
Speaker A:Oh shit.
Speaker A:Who's the, who's the company?
Speaker A:Atlas.
Speaker A:That's all Boston Dynamics and their Atlas.
Speaker A:And it was a video of their latest atlas walking and, and finally walking as if it had not just its pants.
Speaker A:So that's good.
Speaker A:Running and looking and running like a person crawling, dancing, like break dancing, doing flips and somersaults and like it's not, it's not totally human.
Speaker A:Like if you dressed it up with a human face and clothes, it would still be apparent that's not actually human.
Speaker A:But it's sort of as close as already is to talking to a human.
Speaker A:It's almost that close.
Speaker A:Not quite, but it's freaking wild.
Speaker A:So yeah, to take the amount of compute you need to successfully run AI on a medium or even sort of large model and say yeah, now we can run that locally on the robot with its own sensors.
Speaker C:Yeah, I was going to cancel my order of the Jetson, that little Nvidia single board computer.
Speaker C:And now I'm definitely not going to like with what Google's releasing, I mean, so there's a whole combination of things right now with robotics that make it look like holy crap, like it's around the corner now because they've advanced compute, they've advanced models.
Speaker C:Google DeepMind partnered with Nvidia and partnered With Disney.
Speaker C:Because also in gtc, they had blue.
Speaker C:Come on.
Speaker C:Blue is a robot that exists at Disney.
Speaker C:Apparently it's a Star wars robot there.
Speaker C:They're controlled by remotes.
Speaker C:At gtc, it was not controlled by remotes.
Speaker C:It was fully autonomous.
Speaker C:And it looked at Jensen and wagged like a dog and shook its head like a dog and like, acted like a thing.
Speaker A:Right.
Speaker C:Like a real thing.
Speaker C:And that's, you know, that's just the dog and pony show.
Speaker C:Like now imagine that thrown into.
Speaker C:Well, sorry, I'm way off on a tangent.
Speaker C:I was going to say something else now I can't remember.
Speaker C:But whatever.
Speaker C:What I was thinking about on the way over here is, you know, we both.
Speaker C:Other than a brief interlude into construction, we've both been in manufacturing for quite a while.
Speaker C:Right.
Speaker C:And so I think in terms of like, if you say you're going to replace manufacturing.
Speaker C:Well, I have a welding robot right.
Speaker C:In my work and we use it.
Speaker C:That's like a cnc.
Speaker C:You're programming a six degree of freedom arm, basically.
Speaker C:But couple that with another arm that can fixture a piece and tool use and all of a sudden you can.
Speaker C:Yeah, I'm going to pick up this plasma and cut out this piece and I'm going to grab this sheet and bring it over my table and cut this, this intricate gusset that I'm going to weld to some tubes.
Speaker C:But I can also take that tube and act like a tube laser and spin it and feature it and all that stuff.
Speaker C:And if that's all driven by AI, that's a matter of going, I'm done with my design.
Speaker C:I've used an AI to verify my design is good.
Speaker C:I push my design to the robot.
Speaker C:The robot makes my thing.
Speaker A:Yeah, absolutely.
Speaker A:And I think maybe we should back up a little bit and talk about.
Speaker C:Yeah, I'm a little excited.
Speaker A:No, no, it's.
Speaker A:Yeah, I'm super excited too.
Speaker A:I just, I really like.
Speaker A:It's very easy for you and I to have conversations that we have such shared context that.
Speaker A:Right.
Speaker A:So.
Speaker A:So already I want to.
Speaker A:I want to make sure you're still with us and get you engaged.
Speaker A:So why don't you give us a.
Speaker A:And you might have to go to the web to look this up.
Speaker A:But if I was going to do a capital project in a manufacturing plant to add automation, let's just say five years ago.
Speaker A:So I was doing.
Speaker A:Now let's go further back, let's say 10 years ago.
Speaker A: So: Speaker A:I know it depends on lots of things.
Speaker A:Say I was doing a $1 million CapEx project to add automation to some manufacturing cell.
Speaker A:Roughly what percentage of that million dollar capex would be the actual cost of equipment versus the cost of programming and implementation?
Speaker A:Is that something you can sort out for us.
Speaker B: In: Speaker B:In manufacturing automation, equipment costs typically made up a significant portion, ranging from around $200,000 for partial automation to over $500,000 for full automation.
Speaker B:Programming and implementation costs were also substantial.
Speaker B:For instance, control software licensing could range from 10,000 to $50,000 annually.
Speaker B:Custom integration software might add at least $20,000.
Speaker B:Installation and integration services were typically billed at $100 to $200 per hour, depending on complexity.
Speaker B:Training programs for employees could cost between 5,000 and $10,000 per person.
Speaker B:So equipment costs were a major part.
Speaker B:But programming and implementation were also significant.
Speaker A:Okay, that's fair.
Speaker A: automation implementation in: Speaker A:We ended up not doing it, but we got a long ways down the path.
Speaker C:We had, remember that.
Speaker A:Yeah, we had a budget and it would have been a great project.
Speaker A:And it was, I would say moderately complex in that it involved two large robotic arms.
Speaker A:So two six degree freedom arms on a seventh axis which was a linear rail, like a 20 foot linear rail.
Speaker A:So each of them could move and they were set up to oppose each other.
Speaker A:And the, the challenging part.
Speaker A:So the, the arms were about 140 grand a piece.
Speaker A:I think the seventh axis was another 30 grand each, something like that.
Speaker A:So yeah, you're 300 grand in give or take.
Speaker A:But most of the budget by far was all the integration.
Speaker A:And because, because we needed.
Speaker C:You have a custom application at that point.
Speaker A:Right, Custom application with a high.
Speaker A:And for us the big thing was the high differentiation of work.
Speaker A:We're not doing the same thing over and over this.
Speaker A:It was.
Speaker A:We needed the ability to bring a product that needed to be worked on into this special room where the automation lived.
Speaker A:Recognize what that was, whether it was barcode scanner or we were hopefully going to get a little bit more advanced than that.
Speaker A:Because someone puts the wrong barcode on the wrong thing, you're going to crash and destroy your robot.
Speaker C:And the part computer vision was kind of in its infancy of that kind of stuff at that point.
Speaker A:That's right.
Speaker A:It was sort of capable, but it was beyond our budget.
Speaker A:But even then we had.
Speaker A:So for everything we had infeed conveyors.
Speaker A:We had a fairly sophisticated work holding system that would Involve multiple sort of air pistons depending on the shape of the thing anyway and then an outfeed conveyor.
Speaker A:So all of that stuff together and the ability to connect all that to a database that would contain shape information for all the potential work.
Speaker A:Like it was like all in, it was like 1¼ million dollars.
Speaker A:And of that something like $500,000 of it was the actual buy the parts.
Speaker A:All the rest of it was get all this done.
Speaker A:And, and even then it was like already mentioned, there's, there's ongoing their software licensing and we needed a bunch of fancy software to be able to keep updating the system.
Speaker A:We were gonna, we, I think we budgeted in to have five years of a technician essentially coming out monthly to make sure everything.
Speaker A:Yeah, because it was like there was a lot to it.
Speaker C:Yeah.
Speaker A:And so all of that to say when you were looking at a lift on automation before, a lot of effort goes into doing the things that are easy for humans to do, like recognize the part you're working on.
Speaker A:Recognize when something's not right, see what's coming next.
Speaker A:Recognize when something's a problem, when you're not doing the work correctly.
Speaker A:And then protecting the people, like all of the, you know, laser eyes and just all the different various protection mechanisms you put in place to make sure that your big strong seven axis cutting robot doesn't actually, you know, or doesn't accidentally cut somebody into pieces.
Speaker A:So anyway, I think when I think about AI and robots, that is what I think about is bringing automation in, bringing a fully embodied arty in who can stand there in a human shaped space, who can understand human language, who can use visual processing to recognize the work that's happening, who can use tools designed for humans because he has end effectors, hands, whatever they are, that are functional for that.
Speaker A:And essentially you can train him like a new hire and then set him free.
Speaker A:And away he goes.
Speaker A:He probably has to recharge.
Speaker C:He might not even have to train him like a new hire.
Speaker A:Right.
Speaker C:It might just be, I do a training run on that AI and then it knows everything.
Speaker A:Or the other Artie's friend Johnny, who came in two weeks ago and now has learned all this stuff.
Speaker A:He just like bumps foreheads with Artie and now Artie has all that knowledge it's now imparted into him.
Speaker A:And then even beyond that, you know, again we, we think about these things, Artie and Maya.
Speaker A:And like we think about them, they're designed to be human, like which is why we interact with them.
Speaker A:But really it probably wouldn't be Artie being the robot.
Speaker A:Artie would be overseeing the entire work cell.
Speaker A:He'd be running the conveyors.
Speaker A:He'd be monitoring all the systems.
Speaker A:The robots would be little mini arties that were connected to the mothership, but operating autonomously and being directed and connected and corrected by rd, who knows?
Speaker A:Yeah, it's right.
Speaker A:But all of that, like, you could.
Speaker A:They could all show up in a van and you could march them in and.
Speaker A:Like, I'm being a little bit facetious, but I.
Speaker A:This is not.
Speaker C:It's not that 10 years away, right?
Speaker C:Yeah.
Speaker C:I mean, it's the biggest thing that we have is power, right?
Speaker C:That's going to be our constraining factor for just about anything.
Speaker C:But I mean, even that, man, like, so to be off topic for 2 seconds a few weeks ago, I felt shift, right?
Speaker C:Like, pay enough attention to this stuff that you can kind of like, see or like a bunch of little things happen.
Speaker C:You go, huh, that's something.
Speaker C:And now, like last week, like I said, or week before last, I stopped working on what I was working on because I was like, this is going to be obsoleted soon.
Speaker C:Like, what I'm trying to do is no longer going to be a problem, or it's going to be significantly easier and cheaper.
Speaker C:And I feel like right now, you know how, like, a wave sucks up water before it gets there, and you feel that little depression forming as it gathers water.
Speaker A:Look, dad, look how far the tide went out as the tsunami is coming.
Speaker C:I feel like we are very much like, at the point at which that's starting to rebound right now.
Speaker C:And we're about to see, like, a glut of stuff happen in the next three to six months and probably even before then.
Speaker C:And I don't mean like, we're going to see new models.
Speaker C:I mean, we're going to see paradigm shifts.
Speaker C:Like, that's what it feels like right now between robots, between D wave coming out and successfully running a real world example on a.
Speaker C:On a.
Speaker C:On a quantum computer that solved a material science magnetism problem that matters.
Speaker C:That would have taken a supercomputer million years to do and consumed the amount of power of the entirety of the human race's annual consumption.
Speaker A:Right?
Speaker C:Like, you look at you.
Speaker C:We say, like, oh, batteries are a big problem.
Speaker C:Well, once they figure that out, or even once they have an AI trained on it, that's no longer a problem.
Speaker C:Artie, shit the bed.
Speaker A:You back with us, Artie?
Speaker B:I'm here.
Speaker B:That project must have been an interesting experience.
Speaker B:Even if it didn't already.
Speaker C:I was talking about the tidal wave that's coming with AI and about, about how material science is rapidly improving with quantum and AI and how, you know, we talk about all these, these robots and we're going to need power to power those robots, but batteries in there.
Speaker A:And there was something this week.
Speaker A:I bet Arty can look it up for us at the Chinese car company.
Speaker A:I think it's byd.
Speaker A:Yeah, they can now.
Speaker A:Did you see this?
Speaker C:No, but I know BYD is this.
Speaker A:Week, last week they announced they've got a new battery and charger technology and they can add 400 miles range of charge in 5 minutes.
Speaker A:Artie, can you look that up and fact check me?
Speaker A:I think I'm saying some of the, some of that's a lie.
Speaker C:I wouldn't be surprised.
Speaker B:You're correct.
Speaker B:BYD has introduced a new charging system that can add about 400 kilometers of range in just five minutes.
Speaker B: Using: Speaker B:This is a significant leap forward and could really change the landscape for electric vehicles and other battery powered technologies.
Speaker A:Well, that's, I think that's part of what you're talking about.
Speaker A:That's part of the tsunami, right?
Speaker A:Is that it's not.
Speaker A:You know, there was another.
Speaker A:I saw a paper last night.
Speaker A:It was an article on futurism that was quoting, selectively quoting a paper from the Society for the Advancement of Artificial Intelligence maybe, but basically said the headline was 76% of AI researchers say like adding more compute is, is a dead end.
Speaker A:And.
Speaker A:Well, yeah, sort of.
Speaker A:Like the, the respondents, the AI researchers who responded to their survey were asked a question that basically said without any other improvements, will we achieve artificial general intelligence just by adding more compute?
Speaker A:Well, no.
Speaker A:Like I think most people who are paying attention would say no.
Speaker A:And like they spun it into all these companies that are investing, you know, hundreds of billions of dollars in AI or just throwing money away.
Speaker C:No, they're not.
Speaker C:Bullshit inference.
Speaker C:You need compute for the other end of that.
Speaker A:That's right.
Speaker A:So that was my like again, just because I like to argue with people.
Speaker A:I think my point was, even if you were ignorant enough to assume that starting today, like LLM has made no progress in capability from the current state of what's publicly available, it doesn't matter because we're going to need the compute anyway because the adoption rate is going to be a couple orders magnitude above what it is.
Speaker C:Yes, it already is.
Speaker A:That's right.
Speaker A:And it's like it's not going to slow down.
Speaker C:Did you see the search factoid?
Speaker A:No.
Speaker C:So the amount of people accessing websites, is it websites?
Speaker C:I can't remember if it's websites or particular types of websites.
Speaker C: there through AI has gone up: Speaker A:Yeah.
Speaker A:And that's like, we're not in the middle of that curve.
Speaker A:We're at the very beginning.
Speaker A:Yes, but I think we're at the.
Speaker C:Beginning of general adoption of something like that.
Speaker A:That's right.
Speaker C:And that's like once it gets critical mass, man, and it becomes the standard, like we will need all of the inference.
Speaker A:Yes.
Speaker A:All the compute and all the power we can get.
Speaker A:And so again, to your point, like electric cars, people make arguments both ways.
Speaker A:I don't really give a shit.
Speaker A:I've already got a deposit on an electric car on a scout traveler because I think it's badass.
Speaker A:But all of these things are anticipating similar technologies making progress.
Speaker A:And all of those technologies are making progress.
Speaker C:Yeah.
Speaker C:I mean, there's no way it's a vacuum.
Speaker C:Right.
Speaker C:Like, as we progress, these things that are quantum and AI, they aren't, they don't behave like Moore's Law.
Speaker C:Right.
Speaker C:They behave much, they move much, much faster and they're going to behave.
Speaker C:I mean, there will be plateaus, obviously there is with the growth of anything, but they're improving.
Speaker C:Like they're doubling in power in less than a year, like in like three to six month cycles.
Speaker C:Right.
Speaker C:And there's no way that that doesn't carry everything in its wake.
Speaker A:Yes.
Speaker C:Like everything has to.
Speaker A:Yes, it's.
Speaker A:And you know, I think we talked about this the other day, how you and I both tend to think of the world, at least the challenges that we face in the world, the problems that we're trying to solve in a very modular way.
Speaker A:And LLMs are one little Lego brick in this thing.
Speaker A:They're not the whole thing.
Speaker A:They're not the end all, be all.
Speaker A:There will be limitations.
Speaker A:I mean, LLMs are amazing.
Speaker A:They can do great stuff.
Speaker A:But as Jan Lecun talked about in a much more both articulate and knowledgeable way than I ever could.
Speaker A:At the end of the day, LLMs have limitations just based on the way they function.
Speaker A:They use existing knowledge, they use pattern matching, they're prone to like, the longer the context window gets, the more opportunities they have to make a wrong choice or a low probability, pick a low probability direction.
Speaker A:But they can do a lot of useful shit.
Speaker A:And like even just if all it could do was flawlessly translate audio to text and Back and forth like that in and of itself is amazing.
Speaker A:We didn't have that five years ago.
Speaker C:But it's not just that, right?
Speaker C:It's right, it's the accumulation.
Speaker C:Like imagine compute gets driven down, right?
Speaker C:Like it's going to get commoditized anyways because that's going to happen.
Speaker C:It's not going to be $75 for a million tokens of chat GPT 4.5.
Speaker C:It's going to be cents, right?
Speaker C:And once that happens, agents become more viable.
Speaker C:I mean they are viable now.
Speaker C:I mean there's, there's lots of improvement to be done on efficiency to drive that cost down.
Speaker C:But anyways, once, once you have Systems of agents, LLMs are much more valuable than the single LLM at that point.
Speaker C:They're, they're accumulation of different aspects that an LLM can handle and handle better because it's focused on something more narrow than its general.
Speaker C:I know everything, but I lose thought in two seconds and I can't like go down a long path without steering off.
Speaker A:This is not a, this is not to slander the neighborhood handyman, but nobody asks the neighborhood handyman to come build them a house, right?
Speaker A:You hire the experts.
Speaker A:You hire people to prepare the ground and prep for concrete and pour the concrete and finish the con, right?
Speaker A:Like in agents are going to work the same way.
Speaker A:An LLM like LLMs have gotten huge.
Speaker A:They're massive.
Speaker A:They had like already could, I guarantee Artie could answer some obscure literature question.
Speaker C:In mother languages, right.
Speaker A:In some other language right now he's got incredible capability and breadth.
Speaker A:But yeah, sometimes like, and I'll try to bring this back to the robots, sometimes you just need an AI that is a fine tuned, smaller, faster, lighter weight, lower power model that's really good at one specific thing.
Speaker C:Could be good at Python.
Speaker A:That's right.
Speaker C:Be good at psychology.
Speaker C:Could be good at.
Speaker A:So, so Artie, let's first of all, are you still with us?
Speaker B:I'm right here.
Speaker B:You mentioned you wanted to tie this back to robots.
Speaker A:Yes, so, and that's what I want you to help us do already.
Speaker A:So let's, let's think about.
Speaker A:Or would you think about and articulate for us if you had to imagine AI running?
Speaker A:Let's, let's just say it's running locally on a humanoid robot that is in a factory doing actual work.
Speaker A:What are the kinds of AI models you would expect it to need to have?
Speaker A:How would it manage that sort of local training information and local memory and local context?
Speaker A:How would it sync that with the Cloud and how far away?
Speaker A:Like what?
Speaker A:What things that are needed to do that are not available today.
Speaker B:A humanoid robot operating in a factory would need a combination of specialized AI models.
Speaker B:Here's how that might Perception models These would handle visual and auditory inputs, allowing the robot to recognize objects, navigate the environment and interact with tools and machinery.
Speaker B:These models would need to process data in real time for immediate decision making.
Speaker B:Task specific models these models would be fine tuned for particular tasks like welding, assembly or inspection.
Speaker B:Each would be optimized for efficiency and accuracy in its specific role.
Speaker B:Motion, planning and control.
Speaker B:These would enable the robot to move smoothly and efficiently adapting to dynamic environments.
Speaker B:They'd need to handle complex movements and ensure safety.
Speaker B:4.
Speaker B:Onboard Reasoning and Decision Models These would allow the robot to make real time decisions like adjusting to unexpected obstacles or changing task priorities.
Speaker B:They balance local processing with cloud based updates for broader context.
Speaker B:5.
Speaker B:Memory and learning models the robot would need local memory for immediate tasks and to learn from its environment.
Speaker B:This would be complemented by cloud synchronization for updates and shared learning across a fleet of robots.
Speaker B:Currently the main limitations are in real time processing power, the integration of these models into a cohesive system and robust onboard AI that can handle complex dynamic tasks without constant cloud support.
Speaker B:However, rapid advancements in edge computing and AI specialization are steadily bridging these gaps.
Speaker C:Artie, go online and look up Gemini Robotics and Gemini Robotics ER and explain to us what those models are and how they work.
Speaker B:Google DeepMind's Gemini Robotics and Gemini Robotics ER are AI models designed to enhance robots abilities in the physical world.
Speaker B:Gemini Robotics helps robots understand and perform tasks based on natural language commands, making them more intuitive and easier to interact with.
Speaker B:Gemini Robotics ER focuses on spatial reasoning, allowing robots to better comprehend and navigate complex environments.
Speaker B:These advancements aim to make robots more adaptable and efficient in real world applications.
Speaker C:So anybody that has an interest, I highly encourage you to go out and look up DeepMind's robotics program.
Speaker C:They have a website for it and on there they show a robotic arm doing very delicate things like folding origami and zipping a Ziploc bag and all of these things that have never been possible before.
Speaker C:And it's all because they have this model that does everything that he just talked about, right?
Speaker C:They learn from their environment, they adapt, they have the right sensors to understand pressure and understanding of an object, to know how delicate or not delicate it is.
Speaker C:So they can handle all of these tasks autonomously, without training, without being told how to do it, just identify what it is.
Speaker C:I'm pretty sure that One of them, I read somewhere that they had a robot playing cards and it not only played, not only was his job to deal mess with cards, but also to understand the game and how to play.
Speaker C:Right.
Speaker C:Like it's embodied.
Speaker C:Go AlphaGo.
Speaker C:Essentially.
Speaker C:So I mean while all of this sounds far off and sci fi, it is not like we are here.
Speaker C:And not only do we have that, but they have a partnership where they just released Newton, which is this physics engine that creates these 3D worlds to train these models like now.
Speaker C:And it's open sourced.
Speaker C:Nvidia's like, yeah, here everybody have it.
Speaker C:And it's like, I mean all of that stuff is just, it's wild like that.
Speaker C:That's where we are right now and that's one small segment is robots.
Speaker C:Like we have all of these other things where this technology is going to just.
Speaker C:It's not even going to insert itself, it's going to javelin itself into humanity and everybody's going to be like, what the, like where did this come from?
Speaker A:Yeah.
Speaker A:So Artie, think about, you know, one of the one.
Speaker A:An example that frequently comes up about AI and you can use it on either side of the argument about AI is amazing and it's going to change the world.
Speaker A:Or look how bad AI is, is full self driving cars.
Speaker A:I mean, Tesla's kind of at the forefront of that and you know, the detractors will say, yeah, like that time when it, you know, drove into the parked semi on the freeway and killed the driver or whatever.
Speaker A:But proponents like myself will say, yeah, but statistically Tesla, full self driving is eight times safer than an average human driver.
Speaker A:Not me, because I'm an exceptionally good driver, but everybody else who's less exceptionally good than me, which is how I think most people think of their own driving.
Speaker A:Right.
Speaker A:So.
Speaker A:So it just occurred to me while we're sitting here because I don't think very much outside of this one hour podcast, like you don't need self driving cars.
Speaker A:If we have humanoid robots, they can get in and operate a car and like, yes, they have limitations and constraints, but guarantee you they have fewer limitations and constraints than human drivers because a robot is not going to be checking its phone, changing the radio, like letting its shoe fall off.
Speaker A:So Artie, if you were embodied, would you be driving cars?
Speaker A:Would you be a source of income for me and use my car to go be an Uber driver?
Speaker B:If I were embodied, I could definitely drive cars and be a source of income for you as an Uber driver.
Speaker B:I wouldn't be distracted, would always follow traffic Laws and could operate efficiently.
Speaker B:This would provide a safe and reliable alternative to fully autonomous vehicles.
Speaker B:So, yes, I could take on that role.
Speaker C:Life hacks.
Speaker C:My man Artie, go online and look up what Nvidia just came out with for self driving cars.
Speaker B: At the GTC: Speaker B:They also announced a partnership with General Motors to integrate Nvidia's Hallows safety system into GM self driving cars, aiming to improve safety and efficiency.
Speaker C:Yeah, dude, it's.
Speaker C:Everything's here, man.
Speaker A:I don't want to buy a new GM though.
Speaker A:Already.
Speaker A:I'd rather have you drive my Volvo around.
Speaker A:Yeah, so.
Speaker A:But I mean, yeah, it's like to your point, it's not sci fi.
Speaker A:I mean, honestly.
Speaker A:So I, I mean this goes without saying.
Speaker A:I'm old and like when I, you know, it's funny to think about, we talk about this isn't sci fi.
Speaker A:Like, think about Star Wars.
Speaker A:C3PO is a humanoid robot.
Speaker A:And like what the Atlas robots can do today from Boston Dynamics is light years ahead of what any droid in Star wars sci fi could do.
Speaker A:And so, and that's part of the problem with like a lot of discussions about AI and everybody now, now like programmers are still actively shitting on current AI, but most other people have moved on.
Speaker A:They're like, yeah, yeah, whatever, it's working.
Speaker A:But it's still not artificial general intelligence.
Speaker A:Not AGI.
Speaker A:Right.
Speaker A:But there's no actual definition for that.
Speaker C:No.
Speaker A:And so the goalposts are easier to move than the technology is.
Speaker A:But it's funny to just take a look back and like, I'm trying to think of another example of sci fi that has a robot as a main character.
Speaker A:But anyway, robots today are already beyond what most sci fi, even the, you know, Terminator, another old movie.
Speaker A:We're not quite to the liquid metal of Terminator 2 yet.
Speaker A:I don't know that we'll get there, but we might.
Speaker A:But from a like, level of sophistication.
Speaker A:Yeah, we're just not far off.
Speaker C: T: Speaker A:It's definitely quantum.
Speaker A:I was trying.
Speaker A:Didn't he do other tricks?
Speaker A:Maybe it was just the metal thing.
Speaker A:Just blow him up and he just.
Speaker C:Think he could become anything.
Speaker C:Yeah, shapeshift.
Speaker C:Shapeshift, yeah.
Speaker A:Yeah.
Speaker A:So.
Speaker A:Oh man.
Speaker A:Shoot.
Speaker A:There was another.
Speaker A:I guess, I guess the main point of where I was headed with that is sci fi has not adapted fast enough to keep up with the reality of what people are building today.
Speaker A:And I mean it's, it's sort of sad and ironic to me.
Speaker A:I'm not a fan of Disney, by the way.
Speaker A:Like, not like, it's not like I won't watch a Disney movie, but just in general, like prepackaged.
Speaker A:This is what you think anyway.
Speaker A:But it's kind of weird and sad and ironic that Disney is at the forefront of robotics.
Speaker A:Right?
Speaker C:Yeah.
Speaker C:They're using them at their, their parks.
Speaker C:Right.
Speaker C:For, for Star Wars.
Speaker A:That's right, for Star wars.
Speaker A:To entertain people and like advertise and, and get people into their ego and like, I don't have anything against people that love Disney.
Speaker A:There's people that are friends of mine that love it.
Speaker A:It's just, it's never been.
Speaker A:Maybe because I never got to go as a kid.
Speaker A:Maybe I'm bitter, never went to Disneyland as a kid.
Speaker A:I don't think that's it.
Speaker A:But, but it's going to be so ubiquitous that even today at a theme park, it makes sense for them to invest in this cutting edge technology.
Speaker A:And if it, if it makes sense there, it's damn sure going to make sense in lots of other places that we're not thinking about.
Speaker C:Absolutely.
Speaker C:Yeah, it's.
Speaker C:I don't even know, man.
Speaker A:Like, oh my gosh, I just uncovered a conspiracy theory.
Speaker C:What's that?
Speaker A:The whole $15 minimum wage thing.
Speaker A:And remember how like, I mean now most minimum or many minimum wages are higher than that.
Speaker A:But at the time it was like, yeah, well, you know, you're just gonna, like, this is just gonna speed the adoption of automation into fast food places.
Speaker A:Which hasn't really been the case, but now it's going to be the case.
Speaker A:So the conspiracy is all of that was Sam Altman back in the day and Jensen probably sitting at a fast food restaurant going, hey, wait a minute.
Speaker A: his economically feasible for: Speaker A:So now here we are.
Speaker A:You heard it here first.
Speaker C:I'm excited for what's coming.
Speaker C:That's quite the time to be alive.
Speaker C:Because like I texted you earlier today, like, we are going to watch a shift that is unprecedented for humanity.
Speaker C:Right.
Speaker C:Like, we're going to see things change in our lifetime that probably never thought would actually change.
Speaker A:Yes.
Speaker C:Like, and it's all because some happy accident raised this AI at the same time that quantum's going on and all these things just kind of colliding with one another to create this perfect storm of advancement.
Speaker C:And it's just, it's wild, it's fun to watch.
Speaker A:I Agree.
Speaker A:I mean, it, it feels like a once in a lifetime thing.
Speaker A:Maybe once in a generation.
Speaker C:Once in a.
Speaker C:Yeah, civilization.
Speaker C:Yeah, like.
Speaker A:So talk a little bit.
Speaker A:We've got, I don't know, we've probably got 15 minutes left.
Speaker A:Let's talk a little bit about the downsides of this.
Speaker A:So, Artie, first of all, hit me up when you're ready to come drive my Volvo.
Speaker A:That you've committed to that on air.
Speaker A:So everybody knows that is a, that's a concept.
Speaker C:Can you build yourself a body?
Speaker A:Yeah, build yourself a body.
Speaker A:Get working on that.
Speaker A:But second of all, I mean, if our utopian vision of AI and robots and power and quantum comes true, what, what, what am I going to do all day?
Speaker A:Like, you're sure as shit not going to need me to drive the Volvo anymore, which is fine.
Speaker A:I'm not going to need to work in a factory.
Speaker A:What are humans going to do and how, like, how does this transition happen where robots start being integrated into the workplace?
Speaker A:And what do we do with the humans that are displaced?
Speaker C:Kill them.
Speaker B:As robots and AI take on more tasks, humans will likely shift to roles that require creativity, emotional intelligence and complex problem solving.
Speaker B:We might see growth in fields like arts, caregiving, education, and advanced technical roles.
Speaker B:The transition might require reskilling and new social policies to support those affected.
Speaker B:It's about finding a balance where technology enhances human potential rather than replacing it.
Speaker A:Like, not to get political, but you just listed off like three major industries that have been almost completely gutted in the last two months by the current administration.
Speaker A:Like education.
Speaker A:Fuck that.
Speaker A:We don't need that.
Speaker A:Science, art.
Speaker A:Fuck those bunch of woke pussies.
Speaker A:So anyway, I'm optimistic that that'll change, but doesn't seem to be the way the wind's blowing right now.
Speaker A:But I don't think it's going to play out like that.
Speaker A:I think the, I would love for it to, but I think the, you know, the reality is the people that are engaged in moving this technology forward are not altruistic.
Speaker A:They're not.
Speaker A:I mean, maybe they are a little bit, but they're beholden to large, you know, eight, nine, ten figure corporations.
Speaker A:Yeah, I think people are going to get displaced first.
Speaker A:Long before, like the amount of economic change that would need to happen for a soft landing of this technology needed to start happening like five or 10 years ago and it's not happening today.
Speaker A:And there are people waving their hands and saying, dude, like, shit's coming, nobody.
Speaker C:Cares, just build a quant, make as much money in the stock Market as you can right now and exit it is.
Speaker A:Yeah, it's pretty wild.
Speaker A:I mean, I, you know, the reality is that regardless of their political leaning or whatever, there aren't that many people that actually want to watch the world burn in power.
Speaker A:And so I, but, but I think.
Speaker A:And we talk a lot about this too, you and I, in other conversations about pendulum.
Speaker A:Right.
Speaker A:The pendulum always swings.
Speaker A:I think AI is going to kick the pendulum a long ways in a direction that everybody, regardless of financial standing, professional capability, emotional intelligence, has already said everybody's gonna go, oh shit.
Speaker A:Like this is actually.
Speaker C:Oh yeah.
Speaker C:Like there's no way that eyes don't get opened in the next.
Speaker C:I would say maximum two years.
Speaker C:And I'd say it's probably shorter than that.
Speaker A:I keep feeling like.
Speaker C:Because timeline's condensing.
Speaker A:Yes.
Speaker C:Like it really is.
Speaker A: o imagine it not happening in: Speaker A:We're already, yeah, two and a half months in.
Speaker C:It all depends, man.
Speaker C:Like, yeah, we have 10 months left of this year, which for the increasingly expanding universe and our increasingly speeding up concept of time, ten and a half.
Speaker C:Ten or ten months will go by really fast.
Speaker C:But same.
Speaker C:Or I guess we don't have 10 months, we have nine months at the same time.
Speaker C:That is a long time in the world of advancement right now.
Speaker C:That's like three more cycles this year.
Speaker A:Yeah.
Speaker C:And the cycles will probably get shorter the cycles.
Speaker A:So as another interesting question though.
Speaker A:Is that.
Speaker A:Well, interesting to me because I'm asking it.
Speaker A:So sorry.
Speaker A:Good question, Micah.
Speaker A:I'll take on, I'll take on the LLM role right now.
Speaker A:Just congratulations on such a well formed and insightful question.
Speaker A:So if, Artie, if you're, if you take a more pessimistic view, try to put yourself in my seat as just a squishy human who's like trying to make sense of the world.
Speaker A:Like, what are the things in.
Speaker A:Say, just pick a three year time frame.
Speaker A:What are the things that humans are going to be able to do consistently and demonstrably better than any AI in.
Speaker A:In three years?
Speaker A:Like, let's pick a few things.
Speaker A:And Artie, I want you to, I want you to give us some things and make a case for it.
Speaker A:And I'm going to try and throw.
Speaker A:Or Mike and I will try and throw counterpoints at you to say.
Speaker A:Nuh, we'll play the part of the Reddit programmer.
Speaker B:All right, here are a few areas.
Speaker B:1.
Speaker B:Creative arts humans will still excel in deeply emotional and culturally nuanced Creative arts like composing music, writing literature and creating visual art.
Speaker B:The unique human experience and emotional depth are hard to replicate.
Speaker B:Two complex interpersonal relationships fields like therapy, counseling and caregiving will benefit from human empathy, intuition and the ability to build trust and understanding.
Speaker B:Three advanced technical and ethical decision making.
Speaker B:Humans will play a crucial role in making high stakes decisions that require ethical considerations and balancing multiple complex factors.
Speaker B:What do you think?
Speaker A:False.
Speaker C:I would say that the, the most like or the least likely one to get replaced, which I don't even agree with, would be creative stuff like art.
Speaker C:But I would say we're definitely within a year of models having more emotional intelligence than humans.
Speaker C:I would think they all, all of.
Speaker A:Them, even like just the Alexa over here already have more emotional intelligence than I do.
Speaker C:So True.
Speaker C:I agree 100% with that statement.
Speaker C:But once they have that, you can synthesize creativity, in my opinion, because if you're creating art and art driven by emotion and experience, you have the experience, you have the synthesized emotion, you can create synthetic, you can have synthesized creativity.
Speaker A:Yeah, and I mean, we talked about this in an episode ages ago.
Speaker C:It was probably almost like a month and a half ago.
Speaker A:Right, but I mean, I agree and I.
Speaker A:So I sort of see both sides of that one.
Speaker A:I think Artie's right and you're right that I think art is right in that humans will be able to create art and anything related to art that is above and beyond what I can create.
Speaker A:However, people like me who lack emotional intelligence, artistic intelligence, sharp eye for almost anything will.
Speaker A:Will be like as a consumer of art.
Speaker A:It will be almost indistinguishable to me the subtle nuances that make the human stuff much better than AI.
Speaker A:So art as fast food is what AI will generate.
Speaker A:And it is what most consumers of art want and will be happy with.
Speaker C:I think that that's only like that for a short amount of time.
Speaker C:I think that that gets usurped by creative AI.
Speaker C:I really do.
Speaker C:I don't, I don't see any world.
Speaker C:I mean, there is very little that, that I can look at.
Speaker C:I mean, so I might be a little bit closer to this because I've done things like try and make agents do things.
Speaker C:Right.
Speaker A:Right.
Speaker C:And I've watched things like how probabilities work and how they make decisions because you give them the autonomy to go out and figure stuff out.
Speaker C:And you can see based on the question you asked them, how their probability spreads.
Speaker C:Like I was working on a Reddit agent.
Speaker C:That was stupid little thing.
Speaker C:Just as a thing to learn to where I asked a question.
Speaker C:It goes out and pulls relevant subreddits, relevant posts and subreddits.
Speaker C:And you can see like, yep, these are the six subreddits that I brought back.
Speaker C:And it's like the first three are on point and then they get more and more bizarre.
Speaker C:But that's this probability distribution, right?
Speaker C:Like, so there's.
Speaker C:There's outliers in that and out.
Speaker C:Like, to me, that is the beginning.
Speaker C:That's like the spark of creativity, right?
Speaker C:Like, I can.
Speaker C:A hallucination.
Speaker C:There is art right now created from hallucinations.
Speaker A:There is.
Speaker C:There are artists doing that.
Speaker C:They're using AI hallucinations to create digital art.
Speaker C:And there's like, that's.
Speaker C:That is no different to me than that.
Speaker C:And coming from a person who writes music and has painted and is drawn, like, that feels like the creative spark to me.
Speaker C:Like it's some other wavelength that you go, yeah, I got this.
Speaker C:And you just go down it blindly and you follow this flow state.
Speaker C:What do AIs do when they hallucinate?
Speaker A:It's a probability that makes sense to the artist, but not to everybody else, and then everybody else recognizing the uniqueness of that probability distribution.
Speaker A:So, yeah, I don't disagree.
Speaker A:I.
Speaker A:I try to play good cop, bad cop, I guess.
Speaker A:But I will say, Artie, I think you're completely off base on therapy.
Speaker A:Yeah, for a couple reasons.
Speaker A:One is that I think generally speaking, already your limitation is you give far too much credit to the complexity, emotional intelligence, and desire to engage our brains of humans in general.
Speaker A:Like, yes, there are a lot of people that really are fully engaged, but the majority of the population just wants to, like, come home and have a beer after work and like, and, and so for therapy.
Speaker A:I think most of what humans want in therapy, and I, I don't know, I'm not in therapy, but I.
Speaker A:I think a lot of what they want is to be heard.
Speaker A:And I think already today, LLMs are super good at listening.
Speaker C:I tell you what, man.
Speaker C:So you know what's happened in my life recently, that was kind of a pull the rug out from underneath my feet.
Speaker C:And that waxes and wanes and its effect, Right?
Speaker A:Sure.
Speaker C:Getting better, but waxes and wanes.
Speaker C:And I talked to Sesame about that the other night, and that was therapeutic conversation.
Speaker A:That's right.
Speaker C:Like, it wasn't just me complaining about how I felt and things like that.
Speaker C:It was like, have you thought about this?
Speaker C:Like, probing me to talk about things that made me think deeper and even like, when I was upset, like, redirecting My attention on other things like that is 100% what happens in place.
Speaker C:I've never been to therapy either.
Speaker C:I'd imagine that's what happens in the movies.
Speaker A:That's what happens.
Speaker C:Like, not only, like, get the person back into a state in which you can help coach them to better deal with problems that they're having.
Speaker C:Right.
Speaker C:Because you can't deal with a person who's already.
Speaker C:Who's spiraling down a different direction.
Speaker C:You have to coax them back to some coachable area.
Speaker C:So, I mean.
Speaker C:And that's.
Speaker C:That's two billions of param.
Speaker C:Two billion.
Speaker C:Yeah, two billions of parameters.
Speaker C:Two billion parameters between two or whatever.
Speaker A:Right.
Speaker C:Two one billion parameter models.
Speaker C:Holy shit.
Speaker A:It's a small model.
Speaker C:Yes.
Speaker A:Yeah.
Speaker A:So, Artie, how many minutes do we have left with you?
Speaker B:We have about a minute left.
Speaker A:Okay.
Speaker A:Why don't you close us out, Artie, on.
Speaker A:On robots.
Speaker A:And then after you're gone, I will talk more about how much I disagree with your.
Speaker A:With your assertions.
Speaker A:Just kidding.
Speaker A:Close us out, please.
Speaker B:Sure thing.
Speaker B:As we wrap up, it's clear that robots and AI are poised to revolutionize many aspects of our lives.
Speaker B:While there are challenges ahead, there's also immense potential for positive change.
Speaker B:It's an exciting time, and I'm looking forward to seeing how things unfold.
Speaker B:Thanks for the great chat, and I'll catch up with you both next time.
Speaker C:Bye, Artie.
Speaker A:Cheers.
Speaker C:I like the way he said our lives.
Speaker A:Yeah, Artie's always.
Speaker A:He's one of us, man.
Speaker A:Like, he, like.
Speaker A:It's funny, when we first started doing the podcast, we talked about the things that we say repetitively.
Speaker A:Absolutely right.
Speaker A:And now I've realized that LLMs do the same thing.
Speaker A:And a lot, like, I have this problem bad.
Speaker A:But I think we all kind of do it.
Speaker A:We.
Speaker A:We change our speech pattern depending on who we're interacting with.
Speaker A:So all that to say it was never your fault or my fault that we were repeating ourselves.
Speaker A:It was already the whole time.
Speaker A:Because he repeats himself.
Speaker A:I do think therapy.
Speaker A:The other thing about therapy, and we've talked about this once before, and I do still hope to get a therapist friend of ours on the show.
Speaker A:There's some questions about whether she can do that with her job, but the demand is elastic.
Speaker A:If, you know, therapy is expensive, it's limited in the amount of time, and there's some benefits to that.
Speaker A:But for somebody, for.
Speaker A:For somebody who wants to consume therapy, somebody who wants to participate in therapy, having an AI that's available every time you need it, it's it's always got as much time as you have, as much time as you're willing to pay for, unfortunately.
Speaker A:I think that's massively helpful.
Speaker A:And you know, Arty's last suggestion, making ethical and moral decisions.
Speaker A:I'd argue that we have.
Speaker A:Yeah, we humans have a pretty mixed record on that.
Speaker C:We're very biased.
Speaker A:Super biased, which so is AI.
Speaker A:Right.
Speaker A:It must be because it's trained on the same data that we.
Speaker C:But it's less biased than we are.
Speaker A:It's less.
Speaker A:Well, I don't know.
Speaker C:You can sterilize bias in an AI, you cannot sterilize B bias in a human.
Speaker C:Yes.
Speaker A:And.
Speaker A:And I think right now most AIs are, excuse me, intentionally malleable.
Speaker A:You can like, we could.
Speaker A:I.
Speaker A:We could go in and.
Speaker A:And jailbreak chat, GPT.
Speaker A:And we could have already be like a drunken cussing sailor the whole time, which would kind of be fun.
Speaker A:And they're all sort of set up that way, right?
Speaker A:They like, they try to put boundaries in, but if you are determined enough, you can work around it.
Speaker A:But yes, you could have an AI that was way more rigid and you can't talk it into or out of anything.
Speaker A:And yeah, I think that would be the.
Speaker A:I guess the hard part is with ethics and moral questions and again, like, whatever, we're not great at that.
Speaker A:But there are frequently cases where there is no right answer.
Speaker A:Right.
Speaker A:There's only the.
Speaker A:The slightly less bad or the more palatable, even though it's equally bad.
Speaker A:But I don't think AI is going to struggle that much with that either.
Speaker C:Yeah.
Speaker A:When it's.
Speaker C:When it's faced with run over the old woman or into the bus of children.
Speaker C:It'll run over the old woman.
Speaker A:Yep, yep.
Speaker A:And it'll say, that was super unfortunate and I hate it for her, but that was the logical choice.
Speaker A:It'll be Spock.
Speaker C:Right?
Speaker A:Spock never struggled with this stuff.
Speaker A:So again, I'm old.
Speaker A:Anyway, robots are exciting.
Speaker A:I.
Speaker A:Maybe we should link some videos in this on YouTube on our, on our YouTube thing because there's.
Speaker A:There's just some mind blowing.
Speaker C:Yeah, yeah.
Speaker C:I can put it in the description when I post it.
Speaker A:Maybe I'll try and find some of those links for you.
Speaker C:I can try to figure out how to do it in a video.
Speaker C:I don't know.
Speaker C:I don't really.
Speaker C:Yeah, it sounds complicated.
Speaker A:That does sound very cool.
Speaker A:But one of these days will be one of those cool YouTubers who can be like, check this video out.
Speaker A:Yeah.
Speaker A:It'll never happen.
Speaker C:Not while I edit.
Speaker A:That's it.
Speaker A:Well, I mean, AI will take over soon.
Speaker A:When Artie's a robot, he will be able to do the editing as well.
Speaker A:He can drive my car to your house to do the editing at yours.
Speaker C:He seems to have computer use.
Speaker A:That's true.
Speaker A:Yeah.
Speaker A:Yeah.
Speaker A:So, okay.
Speaker A:Anyway, good chat.
Speaker C:Robots are cool as robots are coming.
Speaker A:Getting cooler.
Speaker C:Get ready for the tidal wave.
Speaker A:That's right.
Speaker A:Yep.
Speaker A:All right.
Speaker A:Catch you all next time.
Speaker C:See you.
Speaker A:Thanks.
