Adam Got Suckerberg'd, Humanoid Robots to Do Our Laundry, and Apple AI Sucking
All he fucking did was change his shirt and add a chain. And all of you fucking idiots were like, oh, he's so cool.
Speaker 2:Listen, he did jujitsu and I was a sucker. I fell for the jujitsu.
Speaker 1:He suckerburged you.
Speaker 2:He suckerburged you. I don't know. Did you consider maybe just paying attention to details? I don't know. You're like, here's all the ways that I screwed this up.
Speaker 2:I didn't look at my cable. I didn't plug it in the right port. I don't have that problem. I I looked at all the things, and I made sure I get the full possible bandwidth even if I don't use it.
Speaker 1:Well, I I said in the beginning, I knew when I set it up, I didn't do it right because I had to do it while remember I moved one house over and it was, like, not planned and it was kind of
Speaker 2:sort of rushed. Right. Right. Yeah. Yeah.
Speaker 2:Yeah. It's okay. We can't all be details people, Dax. It's fine. It's fine.
Speaker 2:I mean, the the to be fair, the difference in one gig and 2.5 gigs, not a thing you'll notice probably, unless you really have some reason you want to use that full bandwidth. You know what I mean?
Speaker 1:And and that reason showed up which is why I because like, when I set it up,
Speaker 2:I'm like, okay, I'm I'm gonna
Speaker 1:I know my setup, I'm only getting one gig on each device which is theoretically, that doesn't mean you're not getting the value if the modem is set up correct because you might have multiple devices downloading. Realistically, you
Speaker 2:know Right.
Speaker 1:You're not really gonna have two devices doing one gig each but it's possible.
Speaker 2:It's possible.
Speaker 1:But then I had a new need to be able to download things as fast as possible many times So per
Speaker 2:Oh, man. If we ever get in trouble with the I don't know. Who who read who legislates the the movie watching? Is it the Fed?
Speaker 1:The RIAA or something? I don't know.
Speaker 2:Oh, that's actually a thing? I don't know what
Speaker 1:that is. Okay. The Record Industry Association of America. I guess that's just for music.
Speaker 2:I don't know. Like, rappers can just be like, I shot a guy. He did. Like, if they can say that, why can't we say, I streamed a movie in my house? Like, that seems fine.
Speaker 2:I don't know.
Speaker 1:Yeah, that's true. I just find it funny to be secretive. But yeah, so I wanted to max everything out. But I was surprised, like, it's just so easy to mess it up that every single person in America has a good internet pretty much. Yeah.
Speaker 1:How what percentage of them just have some kind of messed up situation, right? Even the most basic thing of like, your internet speed being good but then like your WiFi router being shitty, Yeah. It's most people probably get a fraction of what they could have.
Speaker 2:Oh, yeah. And they have no idea. I remember like ten years ago when I cared a lot about like my internet quality just starting out working from home, whatever. I I can remember calling whoever it was, CenturyLink or whatever, and being like, hey, I pay for a 100 meg and I'm getting like 12. And then being like, how do you know?
Speaker 2:Like, no one ever calls them and and says that. You know what I mean? Like, just most people just take it for whatever, face value. Like, they get their Internet and they're fine. Yeah.
Speaker 2:But, yeah, when you start actually caring I I started caring about on my local network 10 gig across a few devices. So my my network storage and I have a Mac mini as a server now, as I've mentioned. I really wanted to make sure that was getting the full, you know, 10 gig Do mini does
Speaker 1:the Mac mini have 10 gig? That the new one, right? The new one
Speaker 2:No, this was just an m one. This is the original m series.
Speaker 1:Do you have a 10 gig
Speaker 2:on it? I did. Yeah. You could pay to get like an upgraded 10 gig, Nick, and I did.
Speaker 1:Yeah. Speaking of that, I did upgrade I I did get a 10 gig switch and I can't say the name of the brand because, you know
Speaker 2:Why?
Speaker 1:People might get offended.
Speaker 2:Wait, what? I'm I'm miss I'm forgetting the inside joke.
Speaker 1:I was hoping that you would know what I'm talking about. There's a brand that makes A brand? Really nice 10 gig switches Oh, and their name is kinda weird. I don't know what it is. Their name is Nick Giga.
Speaker 2:Wait. What is it?
Speaker 1:It's Nick Giga, n I c g I g a. Really? It just looks really weird on a box. It's it's probably
Speaker 2:the most extent of how how to explain it. I love it.
Speaker 1:I was so excited when I it was like, Liz, I have
Speaker 2:to show you what I got. And then she was like I
Speaker 1:was like, actually, okay, here's how excited I was. I was she was picking me up because we were going somewhere, so I literally brought the box into the car to show her and she was like, you're so stupid.
Speaker 2:You're such a silly boy. That's so funny.
Speaker 1:Yes. Exactly.
Speaker 2:I I say things like that and there's probably like most of our podcast listeners have no idea about the silly boy thing, and I just sound stupid. But there's like an it's an inside joke, I guess. Well, our inside joke should be the podcast inside jokes. And that one was Liz just said Dax is a silly boy. That's the short version.
Speaker 2:There's a longer version.
Speaker 1:Yeah. And a car full of all of you guys and it really stuck.
Speaker 2:Yeah. It really stuck. So now amongst the terminal crew, that's just silly boy.
Speaker 1:But so yeah. I have this 10 gig setup now. Obviously, the modem is only outputting 2.5 and I don't actually have any 10 gig devices, they're all 2.5. Yeah. But which is, you know, good enough for me for now.
Speaker 2:It's good. I did hit reasons to need 10 gig, which is mostly just the Mac mini reading large files off of the network storage. I think I was using SMB or whatever it is, like the
Speaker 1:I'm using NFS.
Speaker 2:And I am too. There you because chat GPT or something, Claude something told me like we could configure to get more of the bandwidth out of Yeah. Out of that NFS. Anyway, yeah. So I needed the 10 gig bandwidth for that like Mac Mini to read storage fast enough and I went down that rabbit hole, a different one, but somewhere.
Speaker 1:I I actually feel really good because the first time in my life ever, I recycled a device, like
Speaker 2:Like to Apple?
Speaker 1:It's it's No, not to Apple. Instead of just sitting in my closet, I like found a new purpose for it. So my
Speaker 2:Oh, nice.
Speaker 1:My previous Oh, you're laptop so I perfect. Had Yeah. Yeah. My previous laptop I had before this MacBook is now my server, my home server. And laptops make great home servers just because they're so like self contained, like it just I mean, it's it's similar to the Mac mini, I guess.
Speaker 1:And I plugged in a hard drive directly directly to it. So it's now my bootleg NAS.
Speaker 2:A USB? Like an external?
Speaker 1:It's a it's a Thunder it's I got I got a Thunderbolt enclosure for it. Okay. Yeah. But it's just a normal hard. So eventually, if I upgrade to a NAS, I can just put the hard
Speaker 2:drive Yeah.
Speaker 1:Into the NAS. I got it. But right now, it is like the laptop is both the NAS and also the server, so I don't need to read over over network. So it's like accessing the files locally.
Speaker 2:Yeah. Right. Right.
Speaker 1:Right. The performance has been great. Like, it's been great. I solved all of my issues. It's crazy that I can like scrub like a video and like jump to a point and it actually just like loads immediately.
Speaker 1:Like, this just hasn't been a thing that I I've had for a while.
Speaker 2:My mine was all local originally. Like, I didn't go from remote to local, but I had issues with performance and it I didn't realize how sluggish it was until I moved to the Mac Mini. I was using the NOS as the Plex server, which would be great in terms of the bandwidth to disk or the latency to disk, I guess, reading files, but it was not great not being 10 gig over the network or something. There was something about that setup that was not working.
Speaker 1:Yeah. And there's another detail where Intel has something called Intel CPUs have something called quick sync, which is like hardware accelerated video
Speaker 2:Mhmm. Stuff.
Speaker 1:And this laptop happened to have it and it's really good. Like it's it makes a noticeable difference. Not that I'm transcoding that often, but I think it even makes a difference just with any kind of video decoding. Yeah. Big upgrade we've been enjoying and getting full use out of our OLED four k TV and also our projector.
Speaker 2:Yeah. Great. Good times. Home media, Fun stuff. You'd mentioned this giga Nick giga Nick.
Speaker 2:What?
Speaker 1:Nick? Nick Giga.
Speaker 2:Nick Giga, did you hear that, like, TP Link routers sorry. I should never try and say it again. TP Link routers might be banned in The US. There's just just a ton of Chinese stuff. TikTok, TP Link, DGI, like drones.
Speaker 2:I just heard a thing. Just heard a podcast talking about how
Speaker 1:it is. DGI.
Speaker 2:DG DJI. DJ. Did I say g? Yeah. DJI.
Speaker 1:DGA is Directors Guild of America, which is where
Speaker 2:I think you combined the two. Wait, what? Oh, Directors Guild? Oh, yeah.
Speaker 1:That's where our event was. Yeah.
Speaker 2:Yeah. Forgot that's where the the event was. I was like, how do you know what
Speaker 1:the Yeah. The drone thing is crazy. I think I think the next company I start is gonna be in that space.
Speaker 2:I'm sorry. What? You're gonna make a drone company?
Speaker 1:Not exactly. I'm gonna do something a lot more boring. I think drones are super important and will continue to be really really important to be able to manufacture them. Almost all the manufacturing of them is really only in China. Even if you wanna manufacture them here, there are components that drones need that are manufactured in China.
Speaker 1:Yeah. So I wanna see if I can identify one critical component and move it over here. Like I like I think I wanna pair I wanna try to buy like all there's a bunch of manufacturing businesses in America that are run by like someone in their fifties and sixties that are gonna
Speaker 2:Yeah.
Speaker 1:Try to wanna retire in the next ten years.
Speaker 2:Yep.
Speaker 1:I wanna see if I can buy one of those and then repurpose it to manufacture drone parts. That's my next business.
Speaker 2:So this sound sounds hard as a software person. This is all the things that I fear about a business. But it sounds like a good one in light of what I just heard. Have you heard this stuff about DJI? Like how much market share they have in The US and that like American companies just suck compared to them?
Speaker 2:If they do ban them, then like drone is just drone tech in America is just gonna get set back several years and all these companies in The US will have to sprout up and start actually making drones. Yeah. So I think you're you're if you do it now, you're maybe good timing. How do you do that?
Speaker 1:How do you do it now? But
Speaker 2:Yeah. It it probably takes some prep.
Speaker 1:How to do it? I don't know. I've done a lot of things that I just didn't know how to do. I'm just kind of saying, you let go and try to learn it and eventually figure But with this kind of thing, would be you just raise a bunch of money upfront.
Speaker 2:Oh, yeah.
Speaker 1:And then like spend the time trying to acquire the right business and then
Speaker 2:Very capital intensive. Yeah. Okay. Yeah. Can I work at your drone manufacturing drone part manufacturing company?
Speaker 2:Maybe
Speaker 1:if, you know, the location is some place in the Ozarks.
Speaker 2:Okay. Yeah.
Speaker 1:Do you wanna make a website?
Speaker 2:Not like I wanna work on the the factory line or something. Wanna be like a I wanna be the same thing I am now, but just for your drone part company. Don't
Speaker 1:know. Well, idea here is like, you could probably push automation to a crazy degree over time.
Speaker 2:And Mhmm. So I'm
Speaker 1:trying to figure out what that means and what that could look like would be the goal of the company. And if you can figure that out, then you do another part and then another part and then another part and maybe eventually But eventually just
Speaker 2:make the whole drone.
Speaker 1:Yeah. Or like just anything, like any kind of like manufacturing capability means you can make all kinds of things. So I'm kind of very interested in that.
Speaker 2:Okay. So that's a great topic that I have thoughts on. Like, you're saying like a factory with a bunch of machines that are like automating the creation of certain parts. Is that what you mean when you say automation?
Speaker 1:Yes. How can you make it really efficiently? And I I will say like, I think people consistently confuse this detail. China has most of the manufacturing capability. Most of it is manufactured in China.
Speaker 1:That is different than manufacturing efficiency. If America were to make the exact same stuff, we generally do it more efficiently, like with fewer labor inputs.
Speaker 2:Right.
Speaker 1:So I just wanna But it's
Speaker 2:expensive too, right? In America, stuff made in America?
Speaker 1:There's a lot of reasons for that. It could be expensive because our economy has better opportunities for that stuff that the resources that it's required.
Speaker 2:You make products and you need components and those components are more expensive in America, you don't care why, you're still gonna buy them somewhere else. Right?
Speaker 1:Yeah. No. I I I know. But the thing with drones is, it's likely gonna have a military, continue to have a big military use. I'm sure the US government is gonna eventually, if not they're already doing it, stockpile drones.
Speaker 1:Mhmm. Purchasing those components from China is not a good idea.
Speaker 2:No, right. It's a national security risk. That's what I heard on the podcast.
Speaker 1:So you have like a you have a floor of demand that are gonna be it's gonna always gonna be domestic.
Speaker 2:Yeah. So so we jumped off of where I was headed, which was like, I think of like when I remember when Tesla when they were really trying to automate and like scale and they were doing all the stuff with robots in the factories and it didn't go well, like, they ended up
Speaker 1:What do mean it didn't go well?
Speaker 2:Well, like, they hit they had all these growing pains where, like, they tried to automate stuff they shouldn't have and it turned out humans were better at those tasks and they ended up ripping out a bunch of machines that were really expensive. Do do you remember this? Maybe I followed the story more closely.
Speaker 1:I do remember there was a bunch of growing pains like four years ago or or four or five. I remember like a bunch of there was a bunch of things. But like
Speaker 2:So you're saying they figured it out and they ended up in the same place where they're very automated?
Speaker 1:Yeah. It's gonna be some process where you like are gonna do it wrong a bunch and maybe like you Yeah. Went on the right place but just kind of pushing how I'm sure they're they're more and more efficient every single day.
Speaker 2:Yeah.
Speaker 1:To the point where I've seen I don't know if you've seen the Chinese factory like the I think it was BYD, it's like the Chinese Tesla. The inside of their factory looks identical to Tesla. It's like really Oh, really?
Speaker 2:So okay. The still haven't gotten where I was trying to head. Yeah. Maybe two more maybe two more leaps here before we get there. Okay.
Speaker 2:The you've seen these factories where it's these hyper specific machines creating parts and it's this well oiled machine. But like, if they change the product, then they have to like rebuild the factory or whatever.
Speaker 1:Mhmm.
Speaker 2:What I'm hearing all this stuff about the humanoid robots and advancing so much faster. I think you tweeted something about NVIDIA, which I don't know what they did. I'd like to hear from you like what are they doing with robots? Oh, I gotta talk It's about cool. Okay.
Speaker 2:It seems like we're on the cusp of like a bunch of human robots walking around, humanoid, human shaped. And I wonder like, could factories just be a bunch of robots that are human shaped doing the things in a much more flexible way than like these hyper specific machines? Like, is that the future of factories? Is it just a bunch of factory workers with their robots?
Speaker 1:I think sort of. Like, obviously, the goal is to get to a machine that is very general purpose. For a factory, it's probably
Speaker 2:not gonna be a humanoid one. I think
Speaker 1:the humanoid ones are more around getting them into normal society where everything is built for human In my home. Human shit. Yeah. You want one in your home?
Speaker 2:Oh, hell yeah. Are you serious? Casey and I talk about this like every week. Want a human robot. I want a robot that will do all of the things that we have to do every day.
Speaker 1:Yeah. So bad. So there, a human robot makes sense because your house is structured for a human, you have like stairs and and things like that. So if you can move like a human, you can directly do everything. But I think in a factory, it's more it probably I'm sure there's probably some use case for that but what I've seen is more like like a single arm that is very programmable and trainable that can do all kinds of stuff.
Speaker 1:It's not like it doesn't need to be able to, you know, walk around. It can still be stationary. I don't know for the cusp of that. I think there's been this weird you know, how like every industry has like this almost like a benchmark that people are trying to like get to. I think recreating the human gate has been this seemingly simple thing because like, you know, we know how humans walk, like animators can draw a human walking.
Speaker 1:But it's been this weirdly elusive thing where it's been hard to make a robot that really matches the human gate. So that's been like a thing that's constantly pushed. So I think we tend to see like a lot of demos around them walking around Yep. Just because it's like a challenging thing.
Speaker 2:And if
Speaker 1:you can get figure that out, there's probably a lot of other locations there as well.
Speaker 2:And somebody figured it out, right? A Chinese company, I think.
Speaker 1:I did see that video of that Chinese company. That was probably was the best. I don't think it was fake. Okay. There's been a bunch of stuff that I've seen people post that are a 100% fake.
Speaker 1:Even like like Beloji posted one like a year ago that was like a 100% fake and he thought it was real. I don't know if this one's fake, I didn't really look that hard at it. But yeah, it did look really like if that was real, it was walking around in a way that was very convincing. Like a human Why
Speaker 2:do we why do we need them to look like humans when they walk? Is if they could just do all the things that humans do, don't care if they look goofy, right?
Speaker 1:It's not about how they look, it's like if they look correct, that means they're walking the same way a human walks, which means Okay.
Speaker 2:Which is better for
Speaker 1:They can maneuver can maneuver the things we yeah.
Speaker 2:Okay. Yeah. It does seem like this has been a long time coming. I mean, is it DGI? Who is the company?
Speaker 2:No.
Speaker 1:DJ. DJI. It's not it's not
Speaker 2:them, Whatever. But DJ yeah. No. Robotic. What's the name of that company that makes the dog and
Speaker 1:Boston Dynamics?
Speaker 2:Boston Dynamics. I was not close. I was way off. Yeah. Feels like they've been doing stuff for like ten years.
Speaker 2:But like Yeah.
Speaker 1:Where are those robots? Doing backflips and parkour.
Speaker 2:Did they actually sold yeah. Have they ever sold one of those robots to an actual company doing actual
Speaker 1:I'm sure they have. Didn't like Google own them at one point and then Google sold them? So it's been a weird journey for that company. I think Google owned them at one point.
Speaker 2:So you're saying we're not close? I feel like there's so many companies working on humanoid robots that makes me think we're close, but we're not close. I can't buy 1ยข.
Speaker 1:I don't know if we're that close, but let me tell you what NVIDIA has built. And I think this is like one of the coolest things I've ever seen because it combines so many seemingly unrelated technologies into the perfect product in a way that like no other company in the world could even come close to doing. And it's like, I was saying the other day, we're always like, oh, this business, what's their moat? And it's always something stupid like, yeah, our front end has doesn't use React.
Speaker 2:It's just like a dumb like, it's it's always like a little
Speaker 1:like such a small thing. Mhmm. And then when you think about Nvidia and you ask whether remote for the situation, it is like a massive one. So basically, at CES, they announced again, I don't know if this product's actually good. I'm not an expert in this space, like it might end up being stupid, but it was presented in a very compelling way which is they're like, okay, we know all of you are working on robots.
Speaker 1:Home robots, factory robots, self driving cars, they're technically robots. Yeah. We're gonna build you the product you're gonna use to, like, develop this technology no matter who you are. So they have this thing where they had a virtual world, which of course, their history of gaming, they know how to do Oh, yeah. Simulations, simulations, whatever.
Speaker 1:You you can go into VR or whatever and control a virtual model of your robot.
Speaker 2:Okay? Okay.
Speaker 1:So you can control it using motion capture, picking up objects, putting it down, doing a very specific task. Mhmm. In this simulated world where the robot moves the way it would move, where the items move around how real world physics, you know. And it'll record everything you do, right, like data. Mhmm.
Speaker 1:Then it'll use their, just all this generative AI technology to create a thousand x amount of the data as you record it. It'll generate variations of it Synthetic other features. Exactly synthetic data, okay?
Speaker 2:Mhmm.
Speaker 1:Again, running on NVIDIA GPUs in their cloud, all cloud powered. Then they're gonna take that and they're gonna upscale it to be photorealistic. So it's like, it looks like raw video footage. So again, using their GP
Speaker 2:I missed a step. So they have a bunch of motion tracking data.
Speaker 1:In this virtual world, that's like lo fi. It's like,
Speaker 2:In the virtual you world. Yeah. And then they're gonna upscale it to be when you said photorealistic, I'm still imagining like a mesh, like a geometry. But what do you mean by photorealistic?
Speaker 1:Like, it's I mean, this let's say you're like in this virtual factory and you're doing this stuff Yeah. It's not like it's not like somebody spent time to like make everything in a factory look hyper realistic. It's just like
Speaker 2:Right.
Speaker 1:You know, rough shapes of what things look like
Speaker 2:Okay.
Speaker 1:That are accurate but not not don't like if a human looked at a video of it, it wouldn't be like, this is real.
Speaker 2:Yeah, okay.
Speaker 1:So they're gonna take that and they have had this upscaling technology in all of their gaming GPUs for a while where, you know, the GPU renders a shittier thing and then it upscales the frame to be like four k and like nicer
Speaker 2:Okay.
Speaker 1:And more realistic. Yeah. They're basically taking that technology onto synthetic data to create a bunch of, like really much more realistic video footage of these scenarios.
Speaker 2:Okay. Of your product in a virtual space. Okay.
Speaker 1:And like doing the thing with the motion data, all that stuff. Mhmm. Then they're gonna feed that into the training model also powered by NVIDIA GPUs to train the actual AI model that's gonna run on the robot. So it's gonna feed it the video that it would see when it's actually running. But it's like fake, right?
Speaker 1:Was all it's all generated to train the model along with like all other motion data to spit out like a model that you can run on a real robot that could do all of those things. And this whole pipeline doesn't require anything physical. It's all software. You do it all in software.
Speaker 2:Well, okay. I am definitely missing some stuff here. So you don't start the process doesn't start with you physically manipulating objects and doing motion capture on that?
Speaker 1:Sorry. Yes. It it does it does it does it does require. So there's a so my my physical, mean, doesn't involve you physically using the end device or having the end device do anything.
Speaker 2:But like the idea is that video footage just being created is not for you to preview your product in a virtual space. It's the video footage that the product will see because it's as you're building a robot that has a camera. Yeah. Stuff and it's doing stuff. So if you're like iRobot, is that no.
Speaker 2:That's not
Speaker 1:is that a company? Company.
Speaker 2:That's bought
Speaker 1:by Amazon, though.
Speaker 2:Wait. What? Is that a movie?
Speaker 1:It is a movie, but the company the company that makes Roomba is also called iRobot, I think.
Speaker 2:The company and the movie are the same. Oh, I didn't even that didn't occur to me. It's like Will Smith. Right? I really yeah.
Speaker 2:Anyway. So if you're making Roomba vacuums, you would use this to, like, program so that the Roomba, when it sees something, it's smarter about how it acts. Yeah. Yeah. So you
Speaker 1:so this one wouldn't even need anything physical. You would just have this virtual world and you can put obstacles and do all this kind of stuff and then
Speaker 2:Yes. So anybody that knows three d software could build like a a setting, like an environment you wanna build a conquer.
Speaker 1:Three d software is like I think the bar is even lower than that. I think it's meant to be much more accessible than that.
Speaker 2:Yeah. Yeah. But then you'd have to have like a three d version of your product or what?
Speaker 1:Yes. Yes. But what but anyone making anyone building these things already has it.
Speaker 2:Oh, It's like because they have to manufacture it. Okay. Interesting. Yeah. Interesting.
Speaker 2:It's all coming together fine.
Speaker 1:But they they plop that and then you can create all these scenarios, train it in this fake world and then generate what the camera would have seen.
Speaker 2:Interesting. But that means okay. So let's play out the iRobot scenario. The you wanna put pets in your little virtual environment. So you drop a little cat model that you got off of 3dmodels.net or whatever, and you're telling me NVIDIA software is gonna take that cat model and turn it into photorealistic video of a cat.
Speaker 2:Yep. How the fuck? What?
Speaker 1:But we've also we we've already seen this with video generation, right?
Speaker 2:But what if what if it's a really shitty three d model?
Speaker 1:Because it's it's but it's not starting okay. So right now, you can start from nothing and tell an AI model generate me a video of a cat. And it can do like an
Speaker 2:okay That's true. Okay.
Speaker 1:Starting off of a model and upscaling it is like a much easier task and they've already been doing this for years. I don't know if you I guess you have been out of the gaming world but like GPUs, NVIDIA GPUs for a while, the progress in them, most of what they show off isn't raw power. It's all about this trick that they do where, they generate the image from AI. Like what you're looking at isn't like literally what the game told it to render.
Speaker 2:Okay.
Speaker 1:So this is a thing that they've been doing for a while where like they take like a shittier input and make it look a lot better.
Speaker 2:Okay. But like, there's still there's something missing here for me because like, I want to feed photorealistic video to my robot and train it to behave certain way around cats. Okay? What if, like, that requires I put something at least resembling a cat into the environment for it to upscale.
Speaker 1:Right? Yeah. But but this virtual world they have, this is like a whole platform they're building. So again, I don't know if they handle the we we have a drop in cat that walks around and acts like a cat. But you can see how that's actually not that hard because again, they have decades of building virtual worlds and like
Speaker 2:Yeah, know. Okay.
Speaker 1:Like NPCs and all this. This is like a company that knows all of that.
Speaker 2:Yeah yeah yeah. Okay.
Speaker 1:But So but again, the the their demos was more around like manufacturing use cases about like teaching it how to like navigate in a factory floor and like operate on certain items and move things around.
Speaker 2:So okay. So when do I get my robot? Because I really we got a lot of laundry. We got two boys. They're a lot of work.
Speaker 2:I need a humanoid robot walking around bebop and doing whatever we ask it to without saying a word. It's gonna be amazing. I'd pay so much money from one of those right now.
Speaker 1:Okay. So Liz has this theory which I think is correct. All of these companies make what you said, a humanoid looking robot. You're a weird person, so I'm gonna discard your opinion. Ouch.
Speaker 1:Most people that think of this being in their house, they find it from like a gut perspective really creepy because it looks way too much like a person, but it's not really there's a scary energy around it. Sure. So she's like, they just gotta hire Disney and just have them make a cute Yes. Fun robot. And they've already done this.
Speaker 1:If you look at if you I'll try to find the video later, but they've showed these robots that they built that move according to like Disney animation principles. They're they're like cute and they like bounce around and they like do all these like Disney type motions. Uh-huh. Way more likely that that is gonna be
Speaker 2:the type
Speaker 1:of thing that actually gets mass appeal. Yeah.
Speaker 2:I want like c three p o walking around my house doing my laundry. C three
Speaker 1:o just looks like a person.
Speaker 2:That's a Yeah. But like, it wouldn't be creepy because it's like, oh, it's the Star Wars robot. I live in the future.
Speaker 1:It's literally if it's literally c three p
Speaker 2:o It literally looks like
Speaker 1:c three p o.
Speaker 2:I I get the fear, and I will say, if you think about, like, anytime people talk about AI safety and they're like, we gotta make sure that the AI doesn't whatever. We we have to put safeguards because it could just, like, once it's smarter than us, it could do whatever to us. And I always just think, like, just don't let it do something to us. Like, if it's not physical, it's just digital. I guess it could, like, wreck our banking systems.
Speaker 2:It could do all this havoc digitally. But I've never been afraid of AI because it's like, it's not physical. What can it do to me? Then you think about robots and it's like, I really want a robot in my house. Like, I, Adam, genuinely want a robot in my house.
Speaker 2:It's like, oh, this is how the AI could do something to me.
Speaker 1:Okay. So here's a crazier thing. Realistically, v one of these robots won't be fully automated. They'll be remote controlled. Okay?
Speaker 1:So this is now a little bit creepy.
Speaker 2:By who? Wait, what?
Speaker 1:By like a worker and it's like a remote worker, but for your house.
Speaker 2:No. What? I mean,
Speaker 1:a a bunch of the demos you saw like the Tesla robots doing Yeah. Bartending stuff, that was like remote control.
Speaker 2:So if I bought like a $30,000 humanoid robot making up numbers, I would be also buying this person's time to control it around my house when I asked them to do stuff?
Speaker 1:Yeah. And if you think about it, like, okay, if you ignore any gut reaction to it, this is actually totally fine. It's like a pretty good system. You don't need to have a complex AI model. Yeah.
Speaker 1:You just need the physical machinery to be able to do the laundry and someone remotely can like can can control it. Sure. But again, really creepy to have a random person that you've never met in your house, potentially many people on different shifts.
Speaker 2:Yeah. It's like inviting an Uber driver in. It's basically that. Like like having an Uber driver just live with you, some random Uber driver. Because you're putting your life in an Uber driver's hand, it's like similar.
Speaker 1:But here's what really gets me. This would be the first time in history, I would say, where there's like, you're physically in the presence of another person, but there's an asymmetric like violence risk, where they can try to hurt you with zero risk to themselves. Oh. And I feel like fundamentally
Speaker 2:That's not good. Even if the robot
Speaker 1:is slow, even if it whatever, like, this just is a bad system. Because anywhere else, if someone's in front of you, they at least have to worry about their own, like, harm coming to themselves.
Speaker 2:Yeah. Sure. Second book of Three Body Problem, the the restaurant waitress stabbing situation.
Speaker 1:Right.
Speaker 2:Right.
Speaker 1:Yeah. Exactly.
Speaker 2:Yeah. I'm not I'm not letting a robot controlled robot or a remote controlled robot in my home not happening. I want it to be fully AI.
Speaker 1:But how else do we get the training data, right, that would be with this NVIDIA thing?
Speaker 2:Well, I don't know. Why does it need training data? Just make them really really smart and then like, they'll be smarter than humans at some point and then let them in our homes and they can do whatever. Like, I don't know. This doesn't seem hard.
Speaker 2:If ChadGPT can do all the crazy stuff it can do, like, put all that in a robot and let it go. Let it figure it out. Tell it how to move and, like, here's the buttons for moving your arms. It'll figure it out. I don't know.
Speaker 2:It seems easy. So easy to me. At
Speaker 1:that point, what's the purpose of you existing?
Speaker 2:Well, I wanna just enjoy life and I want the robots to do all the tedium.
Speaker 1:Well, it sounds like superior life form. So it shouldn't waste its time doing your laundry.
Speaker 2:You sound like AI. What is the point of you existing? That's what they're gonna say someday. What is the point of
Speaker 1:you existing?
Speaker 2:Right before they terminate us. Oh, man.
Speaker 1:Hopefully, this podcast is not in their training set.
Speaker 2:Yeah. No kidding. Do you ever I hear so many people say, like, that they're nice to chat GBT just in case. Like, do you do you feel that
Speaker 1:nice. I'm I'm really not nice. It pisses me off so much sometimes. I'm just like, shut the fuck up. Stop giving me a full React application when I'm just asking you for an FFmpeg command.
Speaker 2:Honestly, what is the deal with the React bias? It is amazing how much it will spit out React at you if you don't tell it exactly what technology to use.
Speaker 1:But no, it's it's not even that. Okay, like if I'm asking for like, how do I make a button I can click and it spits out React? Okay, that's annoying but I get why that happens. When I say, hey, how do I use FFmpeg to do this? And then it builds me a React UI where I can upload
Speaker 2:Like what? A
Speaker 1:video file to then get it converted. Like, why is it doing that?
Speaker 2:Why? Yeah. I don't know. I guess just like there's an overwhelming amount of GitHub is React apps, I guess.
Speaker 1:Yeah. Yeah. And it's just
Speaker 2:trained on all that. That's crazy.
Speaker 1:TJ just made a video on this I watched this morning.
Speaker 2:I watched it as well. Yeah. Can I be honest? I don't know if Teej listens to podcasts. I love you, Teej.
Speaker 2:This is not a criticism. I couldn't tell if he's memeing. I couldn't tell the tone of his voice in the video the whole time. Just couldn't tell if it's like a whole elaborate troll. Like, he just wanted you to watch a thirteen minute video just to troll you.
Speaker 2:But like, everything he's saying is like makes perfect sense. I didn't
Speaker 1:like you expected. That's it. That's what you
Speaker 2:expect He's from just so funny all the time. He's always joking that it's hard when he's like, I can't tell. It just felt like the tone of his voice like was a big troll. But all the content of it made so much
Speaker 1:That's funny. You're waiting for the punchline. There there were a few jokes in there for sure.
Speaker 2:Yeah. Yeah. Yeah. Would you would you summarize the video for for the audience?
Speaker 1:Yeah. Mean, it's the thing I've been thinking about for a while, especially building a dev tool. His point was if you ask an LM how to do something, it's just gonna reflect whatever current thing is popular. So if you're like Yeah. How do I make a UI where I can do this, it's gonna build you a React app, which is kind of reinforces whatever is currently popular and like puts them even further ahead.
Speaker 1:And so the flavor I was thinking about it is not exactly that. I think about it this other way where so he was describing it as you don't know you're not even trying to evaluate, you're trying to get something done And it's basically doing the evaluation of technologies for you. And the point was that's like gonna lead to a bad place. There's another case where you are evaluating technology. You're like, how should I do CSS?
Speaker 1:And you're picking between Tailwind or something else. But Tailwind has this crazy upside of the training set is massive and your LM experience while using Tailwind is gonna be a thousand x better than anything else. Mhmm. So even if something new emerges that's better, it has to compete with the fact that it's not in the training set and that is like an insurmountable thing. So even someone trying to pick between these two options, they're more likely to pick the one that the LN is good at.
Speaker 1:So it just kind of calcifies everything, makes competition worse.
Speaker 2:Makes sense.
Speaker 1:So yeah, it's like it's not a great situation. I I think about this because SSC is in the training set and Pulumi is in the training set. Mhmm. So at this point, nobody new can really show up because like, you're not doing logic in SST, you're just it's just config, which LLMs are fantastic for. Yeah.
Speaker 1:If you're not already in there, you're kind of it's kind of done at this point. There's not gonna be a new thing.
Speaker 2:Yeah. What's so what's the solution? Or is there one?
Speaker 1:I don't know. It's just an observation that this is what's gonna happen and yeah, like whatever yeah. It's just this thing like like big thing just get bigger and this is just like another
Speaker 2:I guess that's kind of a property of the world as it is. Right? This just exacerbates it?
Speaker 1:Yeah. Big things get bigger but like you need to I think as a society, need to one, allow for things to get big but then also recognize when they're like throttling competition and make it so they can get destroyed. It's just hard to balance all that.
Speaker 2:It it reminds me in the dev scenario, it reminds me of like LSPs. It's like TypeScript gets a lot of momentum because it has a good language server. It's almost like companies that if you come out with a new framework or a new language or whatever, you're gonna have to like your own like LLM workflow. Like Yeah. It's like that's gonna be part of the dev tooling expectation.
Speaker 2:It's like, I wanna be able to build this easily with an LLM telling me exactly what to do or generating a bunch of code. So you have to like deliver that because you can't get it from the off the shelf ones. It's too new and niche.
Speaker 1:Yeah. And it's and like there are solutions for that today, but then then you have the secondary problem of, well, you need to deliver it in a place that isn't open. Like, you need to deliver it Mhmm. In my chat GBT experience. But like, it's kind of hard to do that and kind of annoying to like You can install like a plug in and all, it's just like not Yeah.
Speaker 1:Something you're gonna do. So it's not not great or you're trying to make a new framework? Just give up.
Speaker 2:Well, I actually am trying to make a new framework, so that's Yeah. My
Speaker 1:Guess what? ChatGPT is never gonna spit out your framework.
Speaker 2:It's true. It's not. I'm making a JavaScript framework just so the podcast knows. I said it on Twitter, but that's not one to one with podcast listeners. I'm making a JavaScript framework unironically.
Speaker 2:It is well, maybe ironically. I don't know. I don't know how to frame it.
Speaker 1:No. It's a genuinely good framework.
Speaker 2:It's a great framework.
Speaker 1:It it is the actually, in a lot of ways, it it's not competing with anything. It's the
Speaker 2:only It's not. Option. It's the only Don't think option
Speaker 1:for this type of thing. So
Speaker 2:It's so true. Yeah. I really don't know of any any apps running in this environment. Do you know of any? Is this the first?
Speaker 1:I think so. I've I've I've only seen much more. I've seen cool stuff here, but nothing this cool. Like, it's just
Speaker 2:Yeah. So you'll probably wanna use my framework. You will wanna use my framework if you're building an app for this particular context. Yeah. That's what I'll say.
Speaker 1:The one thing that I say mitigates this a little bit is even let's say take something like Next. Js, which LMs know very well. They also have a problem with this because Next. Js, whatever version number is very different than some other version number. So they're kind of competing against
Speaker 2:I love that you don't know the version numbers. Yeah.
Speaker 1:Yeah. They're kind of competing against their their selves because the LM might be returning the wrong thing. So there is some pressure on the LM to like incorporate new stuff.
Speaker 2:Yeah. Yeah. I do feel like it's not an LLM problem that needs to be solved. If it's gonna if LLMs are gonna be useful for developers, which they are, and there's companies just with that pitch, making their LLM very specifically good for developers. This is like an inherent problem they have to solve.
Speaker 2:Right?
Speaker 1:Yeah. I agree. But like and TJ's another good thing he brought up was to compare it to SEO, which is a great thing to look at because you can see how that product degraded over time. Mhmm. Search is now pretty bad just because there's so many so much incentive to optimize the hell out of it.
Speaker 1:Yeah. Like billions and billions and billions of dollars of incentive and the results are getting worse.
Speaker 2:A great example he said like of them being gamed like the top 10 of Google responses.
Speaker 1:I did tell you about something I learned recently which has been blowing my mind. Okay?
Speaker 2:Mhmm.
Speaker 1:Liz has been consulting at this company. I by I'll I'll say this without specifics. She's been consulting at this company. They're a b to c product. They're a product that they sell to customers.
Speaker 1:Mhmm. They are listed on this Forbes article, saying the top x products in their category.
Speaker 2:Mhmm.
Speaker 1:They're gonna pay them $15,000,000 in 2025 for that Okay. Forbes.
Speaker 2:Yeah. What?
Speaker 1:For that spot. For one article because it generates so much traffic and they're they're like their their customer value is very high.
Speaker 2:So Yeah. But like those lists are just who paid to be on the list.
Speaker 1:In some ways it's bad, but in some ways it also kinda makes sense. So so Forbes has this property, right? They're generate so they're even buying ads just for that article.
Speaker 2:Sure.
Speaker 1:It's like they have this value here. So they are they show up on Google and a lot of their traffic comes through Google Ads, whatever, for this category and people end up on this page. They have 10 slots. The top three slots basically make all the all the money.
Speaker 2:Mhmm.
Speaker 1:They will rotate who is in the top spot based off of, which company performs the best because Forbes only gets paid when someone buys the end product, right? So if the companies are bad, people are not going to buy the end product and Forbes is gonna get less money. Okay. So you only get those top slots if you convert if your product converts really well and then you pay them every time for a conversion. So it's still not a good system but it's not like totally like, oh yeah, anyone can just buy a spot there with a shitty product.
Speaker 2:Sure.
Speaker 1:Yeah. But yeah, is this is this wild? So this one article probably makes like, I don't know, like $3,040,000,000 dollars a year. A single article. Yeah.
Speaker 1:And it's all downstream of Google. And this makes you realize like, no wonder Google makes so much money. This is like like a tiny business inside of the Google machine. And in a lot of ways like, it's like ZigBag is more. This is, it's almost like a parasite because Google is failing to capture this money.
Speaker 1:Like, is capturing it. But Google owns the traffic. They're the one that generate that traffic. And technically, Forbes is making Google search results worse cause they're ranking high for this, but they're basically taking people to an ads page, right?
Speaker 2:Mhmm.
Speaker 1:So again, just a nice demonstration of what a clusterfuck having traffic ends up being. ChatGPT has a lot of traffic. It's gonna turn into some kind of cluster fork like this at some point.
Speaker 2:So I, as someone who is the founder of a startup that very much depends on Google search traffic, I can't listen to this without thinking we should be making more money off of our search traffic. Yeah. We get a lot of inbounds from Google. I'm trying to think what kind of articles could we write that would still rank like our stats pages and who could we sell less than those articles to? That's that's crazy.
Speaker 1:Like you think about these these media companies, like, who the hell, like, they write so much garbage, like, what like what like, how's Ford making any money? But it's because there's these, like, probably handful of properties they have that just generate, 90% of their revenue.
Speaker 2:That's insane.
Speaker 1:That was worth a lot if you can find the right right buyer.
Speaker 2:Yeah. I'm gonna look at those articles totally differently now.
Speaker 1:They do say at the top it's sponsored. Like it's very it's not like a secret thing.
Speaker 2:Yeah. It's not like a shady thing. It's just like it's wild to know the numbers involved. Like $15,000,000 for a spot in an article like that is just mind boggling. What else is going on?
Speaker 2:I feel like there's stuff going on. I there was even stuff I thought of that I wanted to talk to you about like something I saw on Twitter or something. I think there's big news. What's the big news right now? And the TikTok ban.
Speaker 1:TikTok ban.
Speaker 2:Zuck is not cool again. This has occurred Okay. Today.
Speaker 1:I wanna do a rant on this. Perfect.
Speaker 2:Yeah.
Speaker 1:I You've you've known I've been a lifelong Zuckerberg hater.
Speaker 2:Yeah. Yeah. Forever.
Speaker 1:It pissed me off so much when a couple years ago Yeah. All he fucking did was change his shirt and add a chain and all of you fucking idiots were like, oh, he's so cool.
Speaker 2:Listen, he did jujitsu and I was a sucker. I fell for the jujitsu.
Speaker 1:He sucker bird you.
Speaker 2:He sucker bird you.
Speaker 1:It was annoying me so much. I'm like, do you guys not see that this is just that guy in every social circle that just like is just kinda like acting a certain way and not really like doesn't really get it, but he's kinda like copying the things that other people are doing and it's not really
Speaker 2:real. Wasn't just the appearance, it was the lack of like public dumb statements. Like, he kind of like stopped doing a lot of press or something, it felt like. And then like his his actual personality became kind of mysterious because he wasn't like doing a lot of talking, felt like, or at least I wasn't seeing it. Mhmm.
Speaker 2:And then now it's like he can't stop talking. Like the Rogan thing, it's just like, oh, yeah. You're that guy. You're a lizard man. I don't I don't like Zuckerberg.
Speaker 2:I forgot. But I hadn't been exposed to him enough,
Speaker 1:I guess. So I'm glad that it's correcting itself, but it was I was so annoyed when everyone was like, oh,
Speaker 2:Mark Zuckerberg is cool.
Speaker 1:Like, no, he's not guys. He's just lame. Just I
Speaker 2:don't know. And they were doing all this, like, all the open source llama stuff. Like, there were reasons to think like, oh, Facebook might not be the worst. I mean, I would never go to facebook.com, but like, maybe Meta is not the worst. Maybe Zuckerberg's not the worst.
Speaker 2:And then it's all back to where it started and they're the worst.
Speaker 1:Yes. And then, well, here's the worst part. I can't even say anything about it right now because people because right now, he's in trouble for being anti woke. So if I'm like, yeah, Zuckerberg sucks. People are gonna think
Speaker 2:he's Anti woke. Hang on. This is so it takes my brain so long. So woke is if you're very progressive in America, very liberal
Speaker 1:To an annoying degree.
Speaker 2:An annoying degree. That's woke. So if you're anti woke, you're very conservative? Is that accurate?
Speaker 1:No. You're just defining yourself as the opposite of that, so you're equally annoying just in the opposite direction.
Speaker 2:What wait. Wasn't Zuckerberg pretty progressive
Speaker 1:previously? He he doesn't he just he has no real thing.
Speaker 2:He just yeah. He just like puts his finger to the winnings, like, wait, this is my personality now. So
Speaker 1:he's we're watching him do that and then all his stuff about how Apple hasn't done anything in ten years or whatever.
Speaker 2:Oh, yeah. And we're just like, how did he how did
Speaker 1:he step how did he set himself up for that?
Speaker 2:That was like
Speaker 1:That was pretty the easiest no one's in mid or is this that dunk on him? Yeah. Anyway
Speaker 2:Like, if all you think Apple is is the iPhone, then sure, Apple hasn't done anything in a long time. But like if you don't if you're not paying attention to any of the other things like the chips, then sure. I I guess
Speaker 1:Yeah. The chips, which again, go back to the iPhone, which is Mal made this device like competitive with a $3,000 camera, which is totally changing how people like make just the the million different things that you can come up with. Sure. Everyone that's thought that participated in the Zuck is cool thing
Speaker 2:I'm guilty.
Speaker 1:Each one of you, I want you I want you to send me an apology.
Speaker 2:Okay. Well, here's your apology. I'll do it in person face to face. Well, not in person, but on a Zoom call. I'm sorry.
Speaker 1:Here's the other thing that kills me. Okay? A lot of shirts I have kinda look like the shirts he switched to. I also have a gold chain.
Speaker 2:Okay?
Speaker 1:Had one forever. Oh. And here's the kicker. You know, he you know, he does you know, one of the things that he does on his like property in Hawaii? What?
Speaker 1:He's trying to raise the best cattle in the world for the best stakes Oh, word.
Speaker 2:For the best stakes? He's stealing your bit, Dax.
Speaker 1:It's just it's just it's just painful for me. Yeah. And then he has like a like he did that custom Porsche build for his wife. Yeah. That's like the exact silhouette of Porsche he chose was the exact one that I am super into.
Speaker 2:Oh, really? Jeez.
Speaker 1:It's just been it's been hard.
Speaker 2:It's been tough. Well, I'm glad I'm glad he's returned back to the original Zach that we all know and hate. And you can you can have your identity back.
Speaker 1:He's a loser. Nobody forget Nobody forget. He's a loser.
Speaker 2:Yeah. I mean, is the guy that made Hot or Not in college. Like, let's just not forget that he's not special. He is a doofus. That's enough about Zuck.
Speaker 2:There is something cool happening. Nikita launched a new app. How does he keep doing it?
Speaker 1:I left the review for his app.
Speaker 2:Oh, really? He used it? Did it work?
Speaker 1:Okay. Have you tried it?
Speaker 2:No. No. It is the
Speaker 1:most insane onboarding experience ever invented by like a 100 x.
Speaker 2:Like it's really good or it's insane?
Speaker 1:Yeah. It's outrageous. Okay. Okay. So let me the first thing he does, but imagine the app store page, you can probably imagine there's like a picture of the icon, there's a name of the app, there's a manufacturer of like the company that makes the app and there's
Speaker 2:a Mhmm.
Speaker 1:Get button, right? You tap get. He named his company TapGet.
Speaker 2:Wait. What?
Speaker 1:So in the Apple UI, above get, it says tap get.
Speaker 2:Okay. I'm looking for it now in the store. What's what's it called again? Explode. There it is.
Speaker 2:Okay. I've clicked on it in the App Store. Tap git. Wow. That's smart.
Speaker 2:Like to tap
Speaker 1:yeah. If you go through the onboarding experience, it is wild. So there's a part of it because it's kind of like a weird app. You have to like leave the app and set up stuff like in the OS and like your settings
Speaker 2:Uh-huh.
Speaker 1:In the phone.
Speaker 2:Doing it right now.
Speaker 1:When you do that, it uses a picture in picture mode of of iOS to like play a video in the picture in picture mode of what you should be doing. What? Like, he he came up with, every single hack possible. I think it's like he said this he's like he called himself a performance artist. I think this is like performance art.
Speaker 1:It's not like meant to be an app that like is actually used.
Speaker 2:Oh, it's so good. It's like a it's like an experience. Mhmm. Okay. I'm going through it.
Speaker 2:I really have to pee. So this is a great time to take a break. Talk about Terminal Cron membership. And I'm gonna pee, and I'll be right back while I go through the onboarding.
Speaker 1:Alright, everyone. Here's the scheduled ad break. As you know, this podcast is sponsored we have sponsors. We have one sponsor. It's sponsored by terminal dot shop, which is the best coffee if you're a developer.
Speaker 1:If you're not a developer, might not be the best coffee, but if you are a developer, probably the best coffee you can get. So, you know, if you're looking for coffee, s h terminal dot shop. We are about to launch, terminal Cron. Cron is our membership, goes up at your door once a month, a little surprise inside. It's gonna be different every single month.
Speaker 1:You're not gonna wanna miss it. You're gonna see it on all over Twitter, people posting what they got in their cron box. So if you wanna support us, it helps us do wild crazy things including this podcast, consider doing that. So there you go. There's a little ad break while I wait for Adam to finish being.
Speaker 2:Okay. So my review. Are you ready? Yeah. The onboarding is nuts.
Speaker 2:I don't know when I'm in the app and when I'm in something else.
Speaker 1:It's great. Yeah. It's really confusing. I think actually using the app, it's like, again, there's a lot of hacks in there too and I think the experience just doesn't come together that great.
Speaker 2:Yeah. Because I mean, first thing I did was take a picture of my bathroom floor and send it to you, but it just pulls up the normal iOS share thing And I just tapped on your face and it was Slack. And then the message it sent in Slack was Oh, I use the iMessage. Which is the iMessage app. So that is not intuitive, but I get the idea,
Speaker 1:I guess. All the little vibration stuff is really cool. Also, like
Speaker 2:haptic during the I don't if Yeah.
Speaker 1:Like when I send it to you, see how that's like playing?
Speaker 2:Oh, wait because a
Speaker 1:I think that whole thing is a video. And if you click that, like the video opens, so you get this like ugly thing. And then if you tap open, then it actually opens. Meet.
Speaker 2:You would send Meet. That that was a pretty cool experience. I mean, like, just now, that photo you sent.
Speaker 1:I sent you a picture of my Meet.
Speaker 2:I'm gonna send you one just because I wanna actually do it right. Dax, here you are in texts, iMessage. It's very confusing. It's like anyway, maybe we've spent too much time talking about this app because I don't know, you said it's not even serious. Like, this is just a it's performance art.
Speaker 1:I think so. So I have the review saying, Happ is okay, but I want Silicon Valley to get together and agree to never buy one of Nikita's companies or pay him for consulting ever. So he's forced to live up to his potential. Because I feel like like, why is he not trying to be at the level of Zuckerberg?
Speaker 2:Oh, interesting.
Speaker 1:Like he has so much insane talent. I know he just loves the early stage of these apps. Yeah. I'm sure it's a good reason. But I would love to like, if there just wasn't an option to exit, what would he have done?
Speaker 1:I'm like so curious about that. That's a
Speaker 2:great point. Like his apps don't ever continue to exist, do they? Mhmm. They get bought by these big companies and then his talent, the thing he actually made that was so great just goes away.
Speaker 1:Yeah. Like how would he operate
Speaker 2:What in the world?
Speaker 1:Exits weren't an option. Yeah. There's this thing that I think about all the time where I'm like, like obviously, we need a market where exits can happen so that things even get funded. At the same time, I think about Nest and I'm like, Nest was such a good like the Nest thermostat Mhmm. Yeah.
Speaker 1:To this day is still a very good product, even like the current iterations of it. Mhmm. What would it have looked like if they just had to make it? If they just had to make it Yeah. Instead of selling themselves to Google?
Speaker 1:I feel like they would have been if they succeeded, maybe they would have died. But if they succeeded, I feel like they would be like an Apple scale company, but making all this really great home automation stuff and we wouldn't have this clusterfuck of like a thousand options and all of them suck. So sometimes I wonder about these acquisitions, like people don't it gives people such a nice option that they never go for it even if they're capable of it.
Speaker 2:Yeah. There there's a Twitter tweet. Kramer tweeted about acquisition just the other day.
Speaker 1:Yeah. Yeah. Mhmm.
Speaker 2:It was just it's not unrelated to what we're talking about, but just kind of like a frounder perspective take on acquisition and, like, as a failure mode. Yeah. The there's no solution to what you're saying. Right? Like, it's just part it's not like some unnatural legislation that creates this environment.
Speaker 2:It's just normal market forces that lead to those scenarios. Right?
Speaker 1:I would say a little bit. I like, I'm very pro capitalist, but the position that I have, which is different than I would say most people that are pro capitalist, is I think we need more regulation on some of these things that are clearly monopolies.
Speaker 2:Like
Speaker 1:we need to disrupt Facebook and Google. These companies are just like Facebook just buys every single social media app.
Speaker 2:Yeah, right.
Speaker 1:So again, if there was no exit option for WhatsApp, what would they have become? If there was no exit option for Instagram, what would they have become? Would Facebook even exist, you know? So I think, yeah, like the level like these companies have gotten really good at M and A, like they don't let they just don't let themselves be competed with. They take them out pretty early.
Speaker 1:So I just it's down to the founders like having the risk tolerance to be like, no, fuck that. I'm going for all or nothing.
Speaker 2:Mhmm.
Speaker 1:Which like only people like Elon really do. Just like so few people really have that.
Speaker 2:Yeah. So you think about Google, do you do you think like like when you think about the monopoly they've had for what? Twenty years? How long have they had this search and like basically all in traffic? It's like a monopoly on the internet, really.
Speaker 1:Like Yeah.
Speaker 2:The internet flows through Google. Do you think that's gonna like, this is kind of branching off, but do you think that's gonna go away with this whole world of like LLMs and AI and if we re are oh, that we should talk about that. Are we really hitting like some AGI this year that they're talking about?
Speaker 1:Yeah. So I have a few different thoughts on this. So if you remember early early on when we first started talking about podcast when this stuff first came out, I said that I can't see my Google habit going away. I was totally wrong about that. I like do not use Google, I hate using Google.
Speaker 1:It feels so much worse than using ChatGPT. So my use of Google has dropped massively. I'm only there if for a certain number of cases. Increasingly, I've been using Grok for those cases because Grok understands current events, which has been a
Speaker 2:Grok in Twitter. Right? Yeah. Because I just saw yesterday, there's actually a Grok with a q. Website.
Speaker 1:That's two Oh, no. No. No. Yeah. Yeah.
Speaker 1:Grok with a q. Yeah. They're they're like a really fast LM.
Speaker 2:And then there's the Grok website. Is that different from Twitter's Grok?
Speaker 1:No. Grok with a k, they just launched a website.
Speaker 2:Okay.
Speaker 1:Grok.com. Yeah.
Speaker 2:Yeah. I use Grok in Twitter. Just within Yeah. Yeah. I love that they have the button and it'll just like explain a tweet.
Speaker 2:Because there's so many tweets I see and I don't understand them. Even your tweets and I'm like, what in the what are you talking about? And then the Grok will like kinda give me some idea. Usually, it's wrong.
Speaker 1:I've actually never used that feature. Oh, really? Means that my Twitter's too much because I understand everything. No,
Speaker 2:all of it. Yeah.
Speaker 1:But So I was a 100% wrong with that. Like, it totally was a thousand times better and I my cool habit, as strong as it was, has eroded. Mhmm. That said, I feel like with the way these tech companies are, you just never you can just never count them out. They can move like years later and end up winning very silently.
Speaker 1:And the perfect example is Microsoft and Slack. Mhmm. Right? Mhmm. Slack had been around forever, captured this business use case, literally went public and Teams is still bigger.
Speaker 1:Microsoft just moved super late with a worse product and they still won.
Speaker 2:It's just because the inertia, just the world moves so slow that like
Speaker 1:Yeah. The inertia and the the like, the distribution they have and like, it's So I can see Google as the even though it looks like, what the hell, like, what are they doing? Mhmm. I feel like they there's just so much room for moving slow. Like, you you move super slow and doesn't matter, you're a big company, like, you'll still end up winning.
Speaker 1:And Microsoft's, I feel, always ends up winning too in the end.
Speaker 2:Yeah. What I wish Apple would win because I I use Siri because it's convenient, it's on my watch and I hate it. It's so bad in the world Like, of when you're used to getting the kind of crazy stuff we can get out of Claude or or Chad GPT, and then you ask Siri to, like, add something to your grocery list, and it's like, hello. I'm like, oh my god. Like, are they ever gonna get it?
Speaker 2:Are they gonna get the LLM thing on Apple products? Because I would love that. I guess that's what intelligence
Speaker 1:I mean, you can you can now. It's it's just funny. Apple intelligence launched and I, like, don't know what it is still. Like, where is it? Like, I know sometimes I can ask it to ask chat GPT stuff, but, like,
Speaker 2:yeah. I've never even used that. I know when it first came out in the beta, it would summarize text, and that was kind of cool because I do get these free links.
Speaker 1:Still does
Speaker 2:that. It's
Speaker 1:so funny.
Speaker 2:Mine won't do it anymore. And I haven't turned on, and it's driving me crazy because I get nothing out of Apple Intelligence, And I hear about it every day in the news about Apple Intelligence, and it's doing this and it's doing that. And it won't even work on my phone anymore, so I don't know.
Speaker 1:I don't know what they're doing with that stuff. It's it's like comically bad. And I think to be fair, Apple especially does really well when they can wait a very long time. And I don't think that's been a very long time It's that LMs have come like they they put out features like the writing features to help you like generate emails and stuff like that. So that's just like what any random ass startup thinks of
Speaker 2:Right.
Speaker 1:To do. We're gonna generate emails for you. It just it's not that compelling.
Speaker 2:So it it okay. That that you just kind of put like timeline into the conversation. I do remember, I was just thinking the other day, I remember when ChatGPT came out. We were at Reinvent or were you there at that Reinvent? I was at Reinvent.
Speaker 2:I'm pretty sure.
Speaker 1:I don't remember.
Speaker 2:I remember where I was the day ChatGPT came out. I know. Like, it was wild. I mean, was just this kinda like OpenAI was just this research place. And then, like, I remember when this is this new thing.
Speaker 2:You don't remember playing with Chad GBD in the early early days and the excitement around it, how Twitter blew up?
Speaker 1:I just don't really have a clear memory of that.
Speaker 2:Okay. Well, I remember it. And now he what's his name? Sam Sama Sama Yeah. He's saying AGI this year.
Speaker 2:That that's the last thing on the podcast today. What what are your thoughts? What does that even mean? But also, is it gonna happen?
Speaker 1:Sam, like Zuckerberg, I have always gotten a really negative vibe from.
Speaker 2:Me too.
Speaker 1:I I would just look at his face and I'd be like, nope. Okay. He's lying
Speaker 2:to me. Also, like Zuckerberg, there was a little period where I kinda liked him. When he got fired and he was kinda snazz like, sassy on Twitter, I liked Sam for a little bit. I was like, oh, maybe he's not so bad, and now he's back to being Lizard Man. I mean, him and Zuckerberg are, the same human being to me.
Speaker 2:I do not like them. Yeah. I don't even I can't even, like, articulate why. They just seem
Speaker 1:I I think we should listen to these feelings. I think there's some deep ancient wisdom in them, so
Speaker 2:Every I mean, every time Sam says something on like a podcast, he seems so out of touch with like actual human beings in the world.
Speaker 1:I don't believe his intonation of speech is real. To me, it feels like he is doing a very specific way of talking. I don't think that's his real way of talking and that immediately just
Speaker 2:Really?
Speaker 1:Puts me on high guard. Yeah. Yeah. It's it's like, you know, Elizabeth Holmes did this, a lot of people do this, but if you think about again, come back to Elon. Elon does not do this because he sounds so incompetent when he talks.
Speaker 1:He's always stuttering and he's always like interrupting himself and like
Speaker 2:Yeah.
Speaker 1:Yeah. He doesn't sound good. Whereas Sam Altman has this like, he has this like, I'm thinking really hard I thought really hard.
Speaker 2:That is exactly his thought.
Speaker 1:It's like a very practiced calm like Uh-huh. Demeanor. And it's not there's no way you're just talking to your friends like that.
Speaker 2:Yeah. Yeah. Yeah. Well, okay. Regardless of Sam, I shouldn't have framed it with Sam.
Speaker 1:Okay.
Speaker 2:Yeah. Do you the the things are seem to be escalating with AI. It does seem like it's doing things people didn't expect it to be able to do. I even feel like it's doing things I didn't expect to do. Do you know you you have like a disgusted face.
Speaker 2:You don't think so?
Speaker 1:No. It's just here's the thing. It does so many things that are so incredible and useful that it feels like it can do anything. But when you sit down and look at it, it also is like, cannot do I thought I can categorize a wide set of things. Like, it can do this, it can't do that.
Speaker 1:Yeah. I think where it's really impressive and it's I was talking to AJ about the other day, I'm like, we've I've never introspected this hard about what I do all day because like AI is forcing us to think about Mhmm. Like okay, I'm writing code but then like, there's like 12 different versions of that and sometimes I'm
Speaker 2:in this mode and sometimes I'm in that part. Yeah. I can help me
Speaker 1:in this mode, it can't be that. So what it's really good at is having a giant photographic memory. So when you ask it things and it answers
Speaker 2:Mhmm.
Speaker 1:It's really impressive and it's compensating for it's the raw intelligence is pretty stupid. Yeah. That's compensating for the fact that it's never forgotten a fact, basically. So when I'm like, how do I do copy paste using ANSI codes and it spits it out? I'm like, this is amazing.
Speaker 1:I would have never found this otherwise. Right?
Speaker 2:Yeah. Yeah. Yeah.
Speaker 1:When I tell it, hey, I had a scroll back buffer in my multiplexer. Mhmm. When I resize the screen, I need to make sure that things reflow properly. Cannot do it even a little bit
Speaker 2:Mhmm. Mhmm.
Speaker 1:At all. Horrible at it.
Speaker 2:Yeah.
Speaker 1:So clearly, can't compensate for what's required there using memory. You have to, like, have raw, like, logic, some kind of raw horsepower there that it doesn't have. So I don't feel like it's escalating, really. I feel like the way it's being delivered has gotten better, like some the UX things have gotten much better with actually using these things day to day workflows. But it's clear we're just like massaging this fixed amount of raw horsepower into different different ways.
Speaker 2:And you know, it scales like way beyond the current
Speaker 1:I don't know. I mean, that that's a question like every time see something interesting and like I we like to so we do the SC a few times like around AI products. We'll like deep dive into them and just talk about it. We'll do all this complex analysis and like really get down to it. And the answer at the bottom is always the same, which is this is not a good product, but it can be amazing if the raw thing at the bottom gets better.
Speaker 1:It's always that. So the question is always, will the raw horsepower get better? Mhmm. If it can't, none of these products are viable, really. Like, not in like a big way.
Speaker 1:So it just comes down to like, does the scale do the scaling laws hold or is it gonna flatten out?
Speaker 2:I saw a Levels tweet where he launched some like photo or text to video something. And it's like you, it's like your avatar like talking, but it's all based off of like a generated image of you. So it's not don't even have to make video, it's just like you. But it looks so awful, like you'd never wanna like put this video of you talking Yeah. And he acknowledges it and he's basically like, well, well, as the underlying models get better, we can all imagine how this will get better.
Speaker 2:It's like, well, the the that's a big assumption. That has to happen. Like, the underlying models have to actually get better.
Speaker 1:Yeah. So it's it always just comes down to that. Like, none of these products can really make the impact they set out to do today because we're just massaging this this existing thing. Yeah.
Speaker 2:And then there's economics of it. I do wonder, like, are we doing
Speaker 1:That's another crazy disorder.
Speaker 2:Yeah. Are we doing this crazy magic trick, but we're spending way more than we could ever that anyone would ever pay to do this? I think about this every time I hit a limit on Claude. It's like, are we just like, is this all just vapor? It's gonna go away because we can't even afford to do this?
Speaker 2:Like, I would have to pay crazy amounts just for the functionality I get today that wouldn't justify? Like, is it do you feel that way? Like, is this all being mortgaged?
Speaker 1:Yeah. I I I just it's like, I don't even I just want to know what the actual numbers are so I can like understand what types of businesses are possible. I I granted that you can bring the numbers down some big magnitude over time because it is just pure software. But I do remember the days of $4 Uber rides in New York to literally anywhere. Mhmm.
Speaker 1:Mhmm. And now it's like minimum $35 to anywhere. Yeah. Yeah. So I can recall how it feels and how it feels for it to go in.
Speaker 2:They built the habit on a cheap on a cheap dollar amount and then to maintain the habit, it's gonna cost. Yeah. Yeah.
Speaker 1:So like, I just wanna know, like, what are the actual because like, because Sam was saying that, they're losing money on o one pro.
Speaker 2:So I'm like, that is wild because o one pro costs $200 a month. Mhmm.
Speaker 1:Right? Imagine, like, you personally consuming $200 of infrastructure.
Speaker 2:Yeah. It's it's a lot.
Speaker 1:That's insane. Like, no other like, I would say I was trying to compare it to Netflix. I'm like, if you had Netflix on twenty four seven, it's like video four k streaming, that's still probably gonna be like nothing. You know, that's gonna be like like a dollar in infrastructure cost.
Speaker 2:To think
Speaker 1:of $200
Speaker 2:worth of
Speaker 1:Like, what that like, that's that's really mind blowing. Individuals can do that. Yeah. Like a jeep like a like a home GPU costs a thousand dollars. Right.
Speaker 1:So you're telling me I'm consuming that every five months? That that's just like it doesn't it's so crazy. Like, how is that possible?
Speaker 2:That's insane. Okay. I have to unceremoniously get off here. Sorry. This episode is now exploding.
Speaker 2:It's been three seconds or whatever. It's been an
Speaker 1:hour and ten minutes. Yeah. An hour
Speaker 2:and ten minutes. Yeah. Something like that. Yeah. Apparently, we record on Wednesdays.
Speaker 2:That's good to know. I'm sorry. If we always were supposed to and every Wednesday I cancel.
Speaker 1:We we put a I I have a calendar invite. I'm like, okay. Cool. I pay attention to it.
Speaker 2:So Yeah. I do not. I do not pay attention to my calendar. So, sorry.
Speaker 1:Alright.
Speaker 2:Casey also is upset that I don't pay attention to my calendar. This is the thing I should probably start doing to make everyone in my life happy.
Speaker 1:Yep. Anyway Yep.
Speaker 2:See
Speaker 1:you. See you.
