Nvidia’s Insane AI Chip, Apple Gets Serious & Sam Altman Talks GPT-5 | Ep49

KP CUTDOWN

===

Kevin: [00:00:00] Hey, welcome. Welcome. Welcome. Oh, that's you. You do

Gavin: Why is the, now the cameras Totally

Kevin: well, probably cause I was in here. See, he's got to give us, are you, are you mad at our robotic cameraman

Gavin: god damn it get your shit together robotic camera, okay, welcome. Welcome. Welcome. This is AI for humans Your your weekly your weekly guide to the world of generative AI I am here in a weird position if you're not watching the video Kevin and I are actually next to each other on a couch.

Kevin: This isn't AI. I can break the wall. Look at this. We're together. This is uh, Dignation

Gavin: Style. Just as a shout out to our old guest Kevin Rose, Dignation Style. We are in New York City, and we came here. It's weird looking at you while we do this.

Kevin: I don't like it. Yeah. I don't like it. I'd rather be delayed and on Wi Fi and in my own little

Gavin: I have to smell you. I do this, which is

Kevin: Do you want to describe

Gavin: No, that's okay. We'll let the AI, we'll let the AI

Kevin: describe it. Black licorice and leather.

Gavin: Anyway, today, before we jump [00:01:00] into all of this stuff, we want to talk about what's coming up on the show. Kevin, what are we talking about in today's

Kevin: show? So a massive week in AI when we asked the industry to calm down and they knew we sent emails to all the heads of state and they promised no big announcements while we were away and busy with our big presentation. But Nvidia, one of the most powerful companies on the planet had a two hour presentation.

Gavin: they did like an Apple presentation and they kind of mic dropped three different things in

Kevin: it. Like the Taylor Swift of Tech is the Nvidia, CEO new chips. The , compute is going to get crazy fast. The future of robotics.

Gavin: black jacket, though, black jacket, Jensen. We love your leather black jacket. Look, we wanted to get two black jackets for this. Look, we couldn't do it. They were only ones that had giant USA flags at times square downstairs. So we

Kevin: That's

Gavin: do that.

Kevin: So we have all of the news for you. We're going to break it down. We're going to contextualize it. Just the stuff that you need to know.

Kevin: Not necessarily the numbers on the new transistor

Gavin: No, but those are big. Those are very

Kevin: They are big. Plus an apple a day [00:02:00] keeps the doctor away. I

Gavin: Oh no. Is that's worse in person than it is

Kevin: Apple is moving big into AI. I mean, we knew they were going to, but we're finally getting some drippings of information, including some

Gavin: Google,

Kevin: Excuse you. Google and

Gavin: Apple are partnering. Maybe there's a,

Kevin: They make weird noises from the windowsill.

Gavin: there's a rumor that Google and Apple are teaming up, which is pretty big considering they might be losing to Microsoft and open AI.

Gavin: So this is a big deal. And. God knows we need Gemini in our iPhones, right,

Kevin: That's right. We've also got Elon Musk weighing in on all sorts of stuff. And what was the last thing? Oh, Sam

Gavin: Sam Altman.

Kevin: Altman had some updates, new GPT five rumors happening. He's also weighing in on Elon cause you can't. Not disgust him. He's the loudest elephant in the room, despite having one of the weakest AI models.

Kevin: I

Gavin: is kind of like an elephant's name. Elon the Elephant? That would be a really great Yeah, we gotta make a kids book at [00:03:00] some point.

Kevin: I want to point out something. If you are new to this podcast, it is not normally like this. Please go back to other episodes. There's an ounce of dare I say, production value with as rough as it is. Dignation

Gavin: dation did this for 10 years. We can do

Kevin: this.

Kevin: Well, that was 20 years ago. Things have moved on. Point is, this is a very special AI for Humans. I'm delighted to be here with you, Gavin, but, but don't judge us on this presentation just this week. We're doing what we can, , with what we have. Also, every week, this one thing doesn't change.

Gavin: You must subscribe.

Kevin: We beg and

Gavin: subscribe. You must share the podcast. You must tell somebody. Yeah,

Kevin: Do not resist. Do not resist.

Gavin: It's me from the future. I am Gavin Conner, the John Connors least interesting, uh, brother.

Kevin: Am I Kyle Reese or am I a

Gavin: Oh, you're just a person, boring person from

Kevin: the present. Oh, I'm just a background actor from present day.

Kevin: I don't know where John is,

Gavin: to tell you to subscribe to this podcast. So please subscribe to the podcast. Look, we love everybody listens to the show. Thank you so much for [00:04:00] listening to the show. As we always say, people are listening. We have a regular audience, which is amazing.

Gavin: Please leave us a five star review. We will read those at the end of our show. 

Kevin: Whatever platform you're on, leave a comment, engage. Okay, now we hop into it.

Gavin: It's time for the news.

Kevin: Couldn't have said it better myself. 

Kevin: This is the worst thing we've ever done. Please don't judge us

Gavin: Watch it. Definitely get it. If you, even if you're just an audio listener, you should pop onto the YouTube just for this episode, just to see what this adventure

Kevin: is like. This is, this is

Gavin: just, let's be clear, Kevin, you and I are both holding our microphones in our hands for this

Kevin: Him out of frame, it might look a little more professional. That's a

Gavin: That's a good idea. We should just hold them. this is a crazy news week. We did not expect the news to be this crazy. We were always going to do a show, but we expected, Oh, we'll have to kind of come up with a few things, a couple of stories, honestly, the biggest thing that I've seen in a while, really, I Sora 

Gavin: NVIDIA [00:05:00] did a giant two hour live Apple keynote type of show on Monday and kind of blew us away with what they announced

Kevin: Yeah, , they're forecasting and look, they're saying this is a new industrial revolution. Sundar Pichai called AI perhaps the biggest invention since fire. And I'm assuming our audience is familiar with the invention that is fire. , NVIDIA is saying that the future is this AI industrial revolution.

Kevin: And that's convenient because if this is a gold rush, they are selling pickaxes. And what they showed off this week are faster pickaxes that are fully integrated so that if anybody wants to get into AI at the enterprise level, even at the, like the prosumer level, which we'll get to, they have solutions for everyone.

Kevin: Custom hardware. Custom chips, even with robotics. I mean, this was a two hour, he was almost like a rockstar on stage. And , in many ways is in the tech sector.

Gavin: NVIDIA, in case you're just not familiar, you may or may not be. NVIDIA is a [00:06:00] technology company that got famous making graphics cards that power video games for computers and consoles.

Gavin: They have now specialized in graphics cards that really became the main computing source for AI models. So they are probably. The most important cog in the AI production wheel. Their stock is to the moon as they say, because of how much of the AI world depends on Nvidia chips. And Kevin pointed out Jensen Wong is their CEO who, yeah, is a rockstar now.

Gavin: Like, I mean, I think that. In, in the previous years, you had Steve Jobs and you think about the rockstar CEOs. You think of Mark Zuckerberg, you think of people in that level. He's there, which is funny because he's a nerd. He is a, he is an awesome nerd. Like this is a guy that did not come out of a marketing department necessarily.

this is not a concert. You have arrived at a developers conference. 

There will be a lot of science described. Algorithms, computer [00:07:00] architecture, mathematics. 

Gavin: He came up through that and of the two hour presentation, so much of it was like petta flops, uh, this many sequences of cycles of programming power. And it was,

Kevin: not one layer of die on the silicone. Two layers of dye on the silicone and you know, and to a stadium audience packed uproarious applause.

Kevin: And we, after our presentation, we were huddled in a hotel room watching this back like giddy children on Christmas morning because these are future gifts if you believe in the promise that is AI. So look, the too long didn't read the distilled for the broad audience is that they have a new architecture, new chips.

Kevin: And basically you're going to be able to have. Server farms of these chips that are going to crank out and process these AI models so much faster with so much less energy usage, so much less resource usage, 

Gavin: I just think it's weird that I'm sitting here next to you still, but yes, you're right. Absolutely. Kevin, you're absolutely right. The most important [00:08:00] thing I think to take away. So there's a new chip that is called Blackwell and Blackwell is essentially five times more powerful than their last AI chip.

Blackwell is not a chip. Blackwell is the name of a platform. Uh, people think we make GPUs. And we do, but GPUs don't look the way they used to.

This is Hopper. Hopper changed the world. This is Blackwell.

Gavin: And one of the fascinating things about Blackwell is you can tell how big a deal this is. They rolled out quotes from. Mark Zuckerberg, Satya Nadella, CEO of Microsoft, Sam Altman, CEO of OpenAI, Sundar Pichai, CEO of Alphabet and Google, Andy Jassy, President and CEO of Amazon, uh, Michael Dell, Founder of Dell Computers,

Kevin: Dell was in the house, by the way. The camera cut to Dell.

Gavin: Yeah, Demis Hassabis, the co founder of Google DeepMind. These people all gave quotes for this new chip. [00:09:00] Now, when you hear those names, that is the powers that be that we've talked about in AI. So you can tell this is the company that sits at the middle of all of

Kevin: And the Mount Rushmore of AI, all of those faces are on the mountain. NVIDIA is the mountain, right? They're the ones that are making this all possible. And in some ways also trying to compete with NVIDIA down the line, but right now they have all got to play nice. I'm sure the technology is amazing. But as we mentioned on previous podcasts, NVIDIA is sold out of their current hardware.

Kevin: Yes. For like the next two years. And it's all of these massive companies that are their customers buying up every chip they can, because the next future of AI, whether it's video or real time game generation or highly efficient chat bots that can code it's all going to rely on these Nvidia chips to process and crunch all of that

Gavin: So it was two hours long. There was a lot of technical data as, as Kevin, I will say we spun through a lot of it. There was a lot of applause breaks for data numbers, which we always love. We always love shout [00:10:00] out, uh, Jensen

Kevin: Cooling got

Gavin: Yeah,

Kevin: When they, when they showed tubing and like advanced cooling, everybody's like, oh,

Gavin: 3d graphics work on the tubing and server farms that we saw.

Gavin: Okay. So let's talk about a few of the other things besides Blackwell, the new chip. NIMS. Do you want to describe what is a NIM, Kevin? I'm giving it to you because I don't fully think I

Kevin: NIM, well, it stands for , NVIDIA Interface Microservice. So, think of it like a zip file or an executable file on your computer. Think of it like an app on your cell phone, if you will. What NVIDIA wants to do is make a container format For all things AI.

Kevin: So that means if you have a chat bot, right, that's one set of code. That chat bot needs to connect to a model, which is the dataset that powers the chat bot. Well, that's another piece. If you are doing image generation, Gavin, you might have a Laura or something which contains , the character that you want to generate.

Kevin: , you might have an interface that connects to all that. What NVIDIA is doing is offering a container, a module that, you can build for it and pack [00:11:00] All of that information into that one little thing. So that in the future, if you were, I want to integrate with a chat bot, if we want to generate artwork, if we want to create video, rather than having to bolt on three different things, grab the model, get the tool that connects to it, get the thing that lets you customize it.

Kevin: It's all in one. And NVIDIA is even offering basically like a marketplace where you can go and download these individual NIMS that will run on your NVIDIA graphics

Gavin: What's interesting about this , in the presentation, they showed off a bunch of open source versions of these models that were there.

Gavin: Like they had Mixtral plugged into it. They had a couple other things plugged in. And I think the question will be, is will companies like OpenAI allow them to plug things like the GPT 4 API into that? I guess so. Like if the idea is this supposed to be a unified kind of front end to use AI through, and again, the AI, the real money in AI is the pipeline of AI.

Gavin: Maybe they will, but it's gonna be interesting to see how this affects. Cause I think Nvidia, this presentation more than anything showed me, they want to be [00:12:00] much bigger than just chips, right? It feels like they want to own a lot more of the AI conversation, which is in a little bit weird to me about how those old, all those quotes came in because it's all the quotes came in because they knew they, those people all need to buy these chips from open it from a, from Nvidia.

Gavin: But also NVIDIA is going to be a comp, a competitor with them.

Kevin: what I'm saying. Yeah. But if you're, let's say you're open AI or you're meta and you've got your own AI, if NVIDIA gets to scale and they may, because look, they've got graphics cards

Gavin: they got all the

Kevin: in homes where people are going to run this stuff, they'd have a ton of money clearly, and right now they own the compute pipeline for the most part.

Kevin: You're going to have to play nice. You're gonna want your model. , on their cloud and in their infrastructure, because they're going to want it to touch end users. And they're going to want enterprise users to integrate with it. Now, one of the cooler things I think that's easy to grasp, because there were pretty pictures that went along with it is the NVIDIA Omniverse.

Kevin: You want to dive into that?

Gavin: Yeah. So the Omniverse is there really interesting way of creating digital twins of the real world to simulate [00:13:00] environments and allow people to plan out both building plans or robot pathways and warehouses so that you can do something in a simulated universe first and then it would react in a real universe.

Gavin: So basically what you're doing here is you're building an environment that looks like a real world environment. And they showed in the video, they showed like building a giant. Tanker ship or a carrier ship or

Kevin: or a warehouse, which might not sound sexy, but it's fully modeled and has physics. So if you're going to spill a bunch of cardboard boxes into an aisle to simulate a warehouse oopsie, , Amazon can do that virtually and it's super powerful because a, you can test the layout and the structural integrity of the warehouse.

Kevin: Well in advance, you can put your simulated machines in there and see how they were about. You can simulate humans walking around and. Rather than doing this in an actual environment, which would take a lot of time, energy, and money to build, you can do it digitally and simulate millions of hours at the snap of a finger, like one Thanos snap, and you can have tons of data.

Gavin: and you can also, what they showed in their [00:14:00] videos, they showed a way they created a digital SIM version of their giant new server farm for these new chips that they just announced. So they've planned out the entire thing and they know how to best lay it out.

Gavin: And to your point, the robot thing where it showed a robot learning how to go around a problem in the digital SIM. If that way, the robot doesn't have to learn it in real, in the real place. And then all the other robots learn that together. That's how these robots get a lot smarter in the simulated universe, not in the real

Kevin: world. Look, pilots train for thousands of hours in simulators, right? So that they don't have to make the mistake and have the oopsie in the real world.

Kevin: That's what we're doing for, and by the way, for everything. Not just robots as you think, like the bipedal stomping around future Skynet stuff, but like forklifts, cleaning

Gavin: Podcast recordings in hotel rooms.

Kevin: We simulated this and this was the

Gavin: This was better! This was the best version! I'm sorry everybody, this was the best version!

Kevin: So that's the omniverse and look, it has applications for autonomous driving for elder [00:15:00] care, robots for simulating chemicals and even for video games. , one of the coolest demos, cause we're, I think we're going to end up talking about the robots next was, This, , you called it a robo purgatory.

Kevin: That

Gavin: It was so horrifying. I felt so bad

Kevin: these robots. It was robo slip and fall simulator. It's this huge, , 3D environment where they have a bunch of robots learning how to walk and navigate uneven terrain and little

Gavin: Let's be clear. Let's, it isn't just that it is a robot obstacle course designed to get them to fall over.

Gavin: There are little, little up steps and little down steps and they're basically throwing a hundreds and maybe thousands of simulated robots. And watching them fall over. And like, again, this is the stuff that robots are going to look at and say, why did you do this

Kevin: us? There is going to be a class action robot lawsuit in the future.

Kevin: A little robo attorney is going to pop up as this happened to you.

Gavin: Let's, one thing I do want to think about this simulated world is this is the advances that we always talk about. Like, what is the metaverse, right? There is a world that if we start simulating all of the places in the real world, [00:16:00] like this, like if enough of these get kind of connected together, like That is a metaverse,

Kevin: Totally. It's an omniverse, Gavin. You're right. Like the Omnichannel. It's the omniverse. And

Gavin: the way, Jensen Huang, I know we've spent a lot of time on this presentation, but it was a big deal in the AI space. This was in part created for the Apple vision pro, right? So within the Apple vision pro, which, you know, we've talked about in the show quite a bit, there aren't a lot of use cases for yet wearing this and being in the omniverse is a very good use case of the Apple vision pro.

Kevin: And you can pilot yourself around environments and see, one to one at scale, how your environment is playing out. Really, really cool stuff. And keep saying robotics, and that's because the Robits are coming.

Kevin: Yes. Maybe faster than we expected, and NVIDIA is going all in. Well,

Gavin: and so the big thing here is they announced a specific chip, the Jetson Thor Robotics chip for robotics.

Gavin: So, you think AI, you think NVIDIA's already way in advance on stuff. They're seeing the robotic stuff that we've talked about on our show that, that we said [00:17:00] last week. I think there are probably 40 to 50 individual research departments, companies that have these robots that are doing things. They've created a chip for these robots specifically to get better at learning, to get better at knowing how to operate in the real world, to also learn what we've talked about.

Gavin: Tele operating, take a lot of these robots learn by teleoperation, meaning that a human is doing something to show them how to do it, and they learn from that. All of that is going to get better with this specific chip.

Kevin: And it's called Project You say it with me. That was a Jetson Thor. No, the project is group. Oh, Gavin. The project is,

Gavin: Sorry. Sorry. I was

Kevin: confused.

Kevin: The project is group.

Gavin: Groot. Project

Kevin: They called

Gavin: They called it project group. What does Groot stand

Kevin: I think 

Kevin: Generalist?

Gavin: robot, double zero technology. So clearly somebody is a fan of guardians of the galaxy.

Gavin: They wanted to get crude in there. By the way, it's not okay. Well, I'm not even going to, I'm going to argue that. That's a terrible name because Groot is not a robot.

Kevin: No, Groot's a tree. Groot

Gavin: Groot is [00:18:00] a

Kevin: Groot is a

Gavin: Like, that's, it seems like, why not call

Kevin: Because the tree is gonna sprout and then have the branches and Groot will be

Gavin: okay.

Gavin: Okay. 

Kevin: So here's what they're going to do. This Jetson ship that we mentioned, just to round it out. They want it to be everywhere. They want this thing in any robot, whether it's a grabber arm, a bipedal something an autonomous vehicle, your future Nespresso.

Kevin: They want this chip, which has AI learning capabilities, can be easily simulated, they want that chip everywhere, and they are on a path to do that. And there was a shot of Jensen on the stage, Digitally flanked by every robot that

Gavin: was fantastic by the way. It's so cool because by the way, you, you don't realize until you see them all grouped together again, how many people are working on this right now. And again, I keep thinking as a kid that grew up loving science fiction, robots feel insane, right?

Gavin: Like I can't imagine a world where you and I are in this hotel room and instead of us holding these microphones, we have our robot

Kevin: robot arm

Gavin: yeah, the robot,

Kevin: us and doing our camera shot.

Gavin: But that's [00:19:00] not that far away anymore.

Kevin: the beginning of that video, I felt like I was watching the intro to a science fiction movie. It really

Gavin: In a bad way. Some ways it's a little bit

Kevin: scary. It was a little ominous, yeah, it was. But, or they

Gavin: Was it ominous?

Kevin: Versus,

Gavin: Or was it Omniversus? Omniversus. It's the new show coming from Jensen Wong.

Kevin: , that was a two hour presentation. I think we distilled it into two minutes for you. But what you need to know is that the future is coming faster than we expected. NVIDIA wants their tendrils in all of it. And the Robits will be here very, very

Gavin: That's correct. Okay, moving on. We have some big Apple news, Kevin. Because the big rumor

Kevin: in the big Apple, Kevin? Yes,

Gavin: No, not because of that. But thank you so much for that. I really appreciate the attempt to bring that together.

Kevin: I can't strike you through the

Gavin: for

Kevin: sitting next to you. It's time. I'm going to just, I'm going to sit on my

Gavin: we're in the Big

Kevin: my hands. No, we are, we've got some big Apple news. Like everybody knew that Apple

Gavin: Is it because we're in the Big Apple?

Kevin: The city. So windy. , [00:20:00] let's

Gavin: Let's take, let's start over. Let's start over. Okay.

Kevin: Gavin, we have some big Apple news. Is that where you

Gavin: We ask

Kevin: Everybody knows that Apple is moving into AI. If you have an iPhone, you might have heard, like, if you have AirPods in, Apple was describing you photos when people sent them. That was a big surprise, right? Autocorrect seemingly got worse overnight, probably because Apple is testing out new language models.

Kevin: So, Apple's moving into the space, and now they've released , some papers that show that they are training their own language model and also some spy photos of Timmy Cook.

Gavin: Well, let's be clear, this story is twofold. One, there is a bunch of rumors going around that there's a picture of Tim Cook and Google's Sundar Pichai talking to each other in a quiet little restaurant, like a snapped photo from afar. That turns out that photo is years old, right?

Kevin: It is not from today, but the reason it's resurfacing is that there are rumors that I think even Reuters was reporting,

Gavin: It's from Bloomberg news is where the story originally broke that Apple is in talks to let Google Gemini [00:21:00] be their AI model, which listen, I get in that it's a big Google Gemini spent a lot of money and a lot of time on their model. Apple gets a, That's second in class, you know, second in class version of the product.

Kevin: Well, look, Apple also gets billions, literally billions of dollars every year from Google to make sure they're the default search engine.

Kevin: How about a little handshake? If they need a boost on their AI by leveraging Google.

Gavin: the thing I think about this, which is interesting is actually feels like Apple is making a choice in that relationship because Kevin's right. One of the big things people may or may not know is that Google and Apple have had this thing forever. Because so many iPhones exist in the world, Google pays a literal fortune to Apple to be the default search engine.

Gavin: This might be a little bit of a, we think Microsoft might be getting a little too big for its britches. So we go with Google on this deal, even though they're not as good.

Kevin: Interesting. I think I see it slightly differently. Um, that, that could certainly be the case. And we, as we sit here in this [00:22:00] luxurious hotel room, we actually don't know, but if I had to guess, There was some, um, iOS code, which is the operating system that

Gavin: Is that the drug you take in Central America?

Kevin: Yes. It's a ceremonial juice. You sip it and you see the future

Gavin: I've heard people do that and they come back

Kevin: IOSCA. No. Okay. The left drug talk. That's our next podcast. When we flip the table around. , There was iOS code that specifically was sending, , like Siri requests and user requests to a version of GPT, right?

Kevin: So we know that, , Apple has done some sort of deal with open AI or is at least paying to use it. I would imagine a world where Timmy Apple. Doesn't want to show his hand. Like they're all in on GPT, which is still the best model. I think for a lot of use cases. So how do you get maybe a better deal out of open AI and maybe Microsoft who they've also partnered with?

Kevin: You seed some talks like, Hey guys. Maybe we are chatting with Google over here as

Gavin: Or the other thing you in that same vein is you say, Hey Google, we know you're hurting, [00:23:00] kick in extra money and then give us a better deal. And we'll use Google

Kevin: Gemini. Also could be true. And, and this is, by the way, Google's a little bit on their heels.

Kevin: They're a massive company. They're gonna turn out fine, I bet. But they're on their heels a bit with Gemini specifically, because as we've covered in previous episodes, they had a bit of a, , a woke company. gate, if you will. They promised Gemini was going to be crazy, crazy powerful.

Kevin: And while it is a good model, some people were a little unimpressed with what it did. So now would be the time to strike

Gavin: Yeah, exactly. And I think it, who knows really what's going on with what I think is good. All I want is for Siri to be driven by a language model.

Gavin: That's good.

Kevin: anything other than what is driving it

Gavin: because I, Siri is in everybody's pocket, right? And we've talked about PI. We've talked about chat, GPTs, voice app, and everything like that. If Siri is actually driven by a real good LLM, and I would put Gemini in that stage

Kevin: Good enough.

Gavin: that's all I want. So if Apple can do that deal, I'm

Kevin: fine. And look, no company is good with your data. No company is going to guard it perfectly, at

Gavin: Apple's better though,

Kevin: because advertising isn't [00:24:00] their game.

Kevin: Their game is markups and crazy margin , on basic products, , because that's their game. I trust my iDevice more, if you will, whether it's my MacBook or , my iPhone, I trust it to do the compute locally. And so I would probably. be more likely to share my email, my calendar,

Gavin: Yeah. Oh, interesting. I didn't think about that. You're right. You're right.

Kevin: Yeah. So good enough would be way better than

Gavin: Yes, absolutely. All right, let's move on. 

Gavin: , Sam Altman, if you're not familiar, the OpenAI CEO, former and current, , and at one point was no more, but now is the current OpenAI

Kevin: again, off again relationship.

Gavin: The big kind of name in AI went on Lex Friedman's podcast, which is always an interesting place where a lot of news gets dropped.

Gavin: And he had a fun little comment where he talked about GPT 4, his own baby. So let's listen to that.

Kevin: He talks about his own children like my father talks about his children. Gavin.[00:25:00] 

Gavin: Kevin sucks!

Kevin: sucks! Way less enthusiasm from Elder Pereira. Now my son, eh, he sucks. So

Gavin: this was a really interesting interview. I think everybody should go listen to it. Some really big kind of news was

Kevin: dropped. Well, what are the tidbits? Let's

Gavin: Yeah. Okay. So first of all, he said that Sora, his experience of Sora, which obviously he's seen a lot of, he runs the

Kevin: company. Sora is their text to video platform.

Kevin: It's what is responsible for all the videos that are blowing up everybody's timeline. Oh my God. I can't believe it's AI.

Gavin: says he believes that that gives us the understanding that we live in a simulator world because of how well Sora can simulate its world. And if you're not familiar with the simulation theory, go on a Google deep dive, there's a lot of craziness there, but there are legitimately smart people that believe we live in a computer simulation.

Gavin: Like if I punched Kevin really hard on his shoulder, that It doesn't matter because he's not real. I could do whatever I

Kevin: zeros. Yeah, clearly Gavin has never taken my actual human emotions into account.

Gavin: .

Gavin: also, so what else, what else does he have? I'm going to have to say Kevin's

Kevin: Kevin? I don't know. I'm concussed, [00:26:00] Gavin. , if you've ever used OpenAI, now some people, by the way, if you're listening to this and you're like, I've used GPT, I've gone and I've chatted with the Robit and I was unimpressed.

Kevin: Chances are you didn't put money into the AI vending machine to get access to the better model. ChatGPT 4 is amazing. , everybody uses it. It's really best in class. There are other competitors now, but it's the best. People are waiting on GPT 5 and there's been rumors that it's coming out this year. Sam confirmed that they're going to release something new this year.

Kevin: He didn't say if it will be GPT

Gavin: know what he was going to call it.

Gavin: So that may or may not be GPT 5. It could

Kevin: be something new. I wouldn't mind if they rebranded it, honestly. Sure. Because saying GPT this and GPT

Gavin: they call it? Anything but GPT. Like GP tot? Turtle? GP

Kevin: ha ha ha! Just get away from the GPT. It's a,

Gavin: The, uh, uh, Open AI iverse? Open AI iverse? Part one.

Kevin: How is it that you got to hit me? Ha ha ha! But I don't get to hit

Gavin: [00:27:00] because I had the line about the simulation.

Gavin: I got this. I said it, I set it

Kevin: it. Okay, maybe we can punch up his copy.

Gavin: Okay. Fair enough. That GPT,

Kevin: they're releasing something this year, but here was the interesting part is that when specifically talking about how much better the next model will be, Sam said, and he's not one to typically hype this sort

Gavin: he really

Kevin: He really isn't. But here's what he said. 

Kevin: The difference between GPT 3 and GPT 4, that's the delta that will be the difference between GPT [00:28:00] 4 and whatever they release next. And I'm telling you, if you go back and try to have any sort of a conversation or get GPT 3 to do anything for you, probably not going to

Gavin: do it. We use GPT-3 a lot and by the way, we're not saying the original chat GPT, 'cause the original chat GPT was GGPT 3.5.

Gavin: This is GPT-3 0.3 or GPT-3, which is

Kevin: broad audience right now, they're like, what are you saying? And and that's fair. That's fair. The point is, the next version is not as Sam is, is alluding to.

Kevin: Yes. It's not gonna be a little bit better. Yeah. We're talking demonstrably better. Yes, markedly better. So is it going to help you code or write or generate artwork or go surf the web and give you notes or who knows? Yeah. It's just going to be that much better and I, again, Sam, I don't shade him or OpenAI.

Kevin: So for him to hype this, I think is a big deal.

Gavin: And the last thing Sam said, he had some kind words or not so kind words for a buddy, old pal, Elon, the former co founder of OpenAI.

Gavin: He said that Elon's lawsuit, because [00:29:00] Elon is suing OpenAI, says is unbecoming of a great builder. And it makes him sad. He wants the old Elon

Kevin: back. We miss old

Gavin: We miss old Elon.

Kevin: miss old Elon. And look, I thought like people are waiting for Sam to say something. They're waiting from some, for some statement. This is pitch perfect, right? Cause it, it says, Hey, Elon, I'm not mad. I'm disappointed.

Gavin: know what I have to say about Sam? Sam, I feel like is the first tech giant who has come up in the clapback social media space in that he knows how to play the social media game. And he's not trying too hard. Elon often, sorry, Elon, you know, you are what you are, but you are trying too hard.

Gavin: Sam knows and grew up doing this. He is somebody that you can tell is a digital snark native and it's impressive. I always loved that part about

Kevin: And take it from us, when it comes to trying hard, we

Gavin: Oh, we are hard tryers. This would be a good chance for you to punch me if you want.

Kevin: I don't

Gavin: Yeah. You missed it. So you miss it. You miss punching me. You missed

Kevin: [00:30:00] this is what the audience doesn't see all the time.

Kevin: I like you.

Gavin: you too. Punching people. This is the way it works in, in this is primitive, man. I've been listening to, I've been listening to a lot of, uh, entertain. Yeah. I've been listening to Joe Rogan. Is that why you

Kevin: in swinging

Gavin: Yeah, I just hopped up

Kevin: hopped up on nootropics right now, alright? He is on it. 

Kevin: Let's hop into it. 'cause we're still talking about Elon because despite our best efforts, he does take up a lot of space. Yes. He is a bit of a digital vacuum.

Kevin: And actually there's, he had a story this week, which I found really interesting. Maybe this is a little bit of the old Elon coming out.

Gavin: Elon says that we are going to need high universal basic income, meaning not just like 10k to everybody per year, which is not enough to live on. Elon is saying that we are going to need to figure out, as a society, a way to pay everybody an actual wage as UBI. Now, If you're not familiar with UBI, you probably are if you're listening to this podcast, but if you're newish [00:31:00] to this, it basically is the idea that in the future, AI will take a lot of people's jobs, but we'll create a massive amount of value to the world that there will be so much surplus value that we will be able to afford to pay a significant wage to people just for living.

Kevin: a universal basic income, because you exist in a world now where there is abundance or abundance can be created essentially on demand.

Kevin: And so, Hey, we're going to need someone to buy the things and use the widgets and exist amongst our

Gavin: And to make choices,

Kevin: Right? So here you

Gavin: yeah. So, and I think his thought here is that he believes there's going to be a lot of jobs phased out, which we talk about. And I think you and I will kind of agree with this is that probably in the. Mid to long term we definitely need some

Kevin: version of this.

Gavin: what it looks like my pessimistic view of this and is in America I am very worried about this ever happening

Kevin: Bucks for

Gavin: No Patriot Bucks for me because I don't believe you would get both sides of the aisle to agree on what Patriot Bucks

Kevin: are. I actually have a [00:32:00] simulation running right now in the

Gavin: Oh How'd it

Kevin: go? It's AI bootstraps. Oh, yeah? And so, rather than actually, like, make sure people have food and shelter in this

Gavin: You can pull us up by

Kevin: environment, Yeah, we're just gonna hand out digital bootstraps so your avatar

Gavin: they have a picture of my face on them. Who's face do I get Patriot bucks

Kevin: They're bespoke.

Gavin: Yeah.

Kevin: Everybody gets their own face on their own

Gavin: So I don't doubt that say there's a lot of people in the world who have talked about ways to give people back stuff. Right. Denmark is a country where there are things like very high paid healthcare.

Gavin: There is some version of this. They give out money for people for doing creative work. Those kinds of countries, I could see integrating this in some form. I worry about America. I worry about the larger democracies trying to figure out a way to make this make it, to make it real. Do you know what I mean?

Kevin: I think the real issue is that Elon probably used a fungi identifying AI app and thought it was nibbling on a portabella and had a transcendent experience,

Gavin: That's probably right. As Kevin just mentioned, we moving on to our next question. story, which is, , a great article [00:33:00] from the Washington Post the headline is using AI to spot edible mushrooms could kill you, which yes, it can, and please do not. So the story here is that, yes. There are a number of different apps that allow people to identify mushrooms. And by the way, this is an amazing subculture.

Gavin: Have you ever dug into people who do this? It's fascinating,

Kevin: have a family member that does it.

Kevin: Yeah. And they gave us a book on it. Yeah. And I'm like, I do not trust myself to make this decision. Exactly. Does this look like the one on the page? Yeah, eh.

Gavin: Well, and the funny thing with mushrooms is you could either die, have an enjoyable meal, or go on a trip and fry your brain.

Kevin: ego could have a death and wouldn't that be enjoyable?

Kevin: Been there. Mushrooms are weird.

Gavin: Anyway, what this is, is a great article that just 

Kevin: reiterates

Gavin: stuff that we've talked about again and again.

Gavin: Do not trust AI to do things that are super specific and could kill you. Because in this article, the funny thing is they talk about a guy who goes out, uses a AI app or an app to identify mushrooms that uses AI. And it identified something as not the thing it was. And he got sick. Right. And this is like something that could really [00:34:00] happen to everybody.

Gavin: But what if this whole story was his excuse to , his significant other that he actually took these on purpose, but he's like, Hey honey, the, the AI app told me this was something

Kevin: Gavin, that's Yelp. What are you doing? I pointed the camera at it, it told me to eat!

Gavin: It wasn't the case, but that would be

Kevin: funny. I like that story, I like that alt version of it. Look, I don't trust AI to write our newsletter for us. I still do that by hand. So don't trust it with the fungi that you're going to ingest while you're out foraging.

Kevin: That's, that's the story

Gavin: Shout out to the Washington Post too, because they're doing a great job of covering some of these stories of AI gone wrong, which we think is really important in this space.

Gavin: Right? Like you cannot use AI for everything. Remember that this is not a panacea. It will not solve all our problems. We really have to keep an eye out as we

Kevin: along. Yeah, so remember, even with Panacea, eventually all those, all those countries and continents drifted

Gavin: That's Pangea, Kevin. That's Pangea.

Gavin: We have a fun little bit going on here, don't we? Would you like me to punch you again? We need to be separated

Kevin: need to be separated [00:35:00] by webcams.

Kevin: Everybody is realizing this now. I feel like I'm commenting on an eSports match. Like, I'm amped up in a way that I don't need to be. And I'm here and I'm

Gavin: We gotta be a little

Kevin: they've got a dragon in the woods. We've got to put down some

Gavin: We should do a bit sometime where we an AI e sports match and we comment

Kevin: on it.

Kevin: Oh, I was, okay.

Gavin: Or we have an AI e sports commenter as our co host. Yeah, yeah, yeah. That's a good idea too. Okay. It's now time for AI.

Kevin: I see what you did

Gavin: See what you did there.

Gavin: This is our segment every week where we look at some fun things about AI in the, in the universe of the world. That's right. I'm so tired.

Kevin: the omniverse of the everything. Here's the thing. Okay. Every week, papers come out, videos are released. It's stuff that maybe you and I can't actually run [00:36:00] yet. They haven't released the code, or maybe it's teasing something that isn't even out yet. This is our little section to say, hey, I see what you did there.

Kevin: So what do we got this week,

Gavin: Okay, so we're first talking about Google Vlogger. This is a very interesting, cool thing.

Gavin: , you think about Gemini and all the stuff that Google does, but they also are a huge research department, and they release stuff all the time. Google Vlogger, which is a hilarious name to me because it is such an old fashioned word. We just talked about this in our presentation today. But it is a thing that allows people to use photos to bring them to life.

Kevin: A single photo.

Gavin: photo. Yes. And you can use it to basically do multiple languages. You can use, uh, have people talk is very much like what we've seen. Hey, Jen do with their avatar thing, but this is a one photo solution, right?

Kevin: If you're just tuning the audio podcast. Trust us. It's realistic looking photos that are moving about, but you should go fire up the YouTubes and take a

Gavin: look. Trust us. That's, that's our new sign out for everyone. Just trust us.

Kevin: trust these guys?

Gavin: Uh, all right, Kev, what about, uh, the next thing up here?

Kevin: So Magnific is an app that we've covered before. Shout outs to a Javi Lopez, uh, who's done a great job of releasing [00:37:00] bunch of AI tools that can enhance imagery. Now Magnific in the past, we've talked about them because it can enhance an AI image.

Kevin: You can get something out of a mid journey or a dolly and then add crazy. , detail to it. You can really, really enhance a photo using AI. Well, now they have a style transfer. So you could take a photo of say two podcasters in sad hotel and tell it to make it look like an anime or make it look like it was shot on a specific type of film or make it look like, , an impressionistic painting and it's, the results are really

Gavin: And here's the thing about Magnific. Javi is great and they packaged this thing together. Now, could you do all of this stuff open source and free? Yes. And we've talked about versions of this. The real benefit here is it's all in one package. Right. And , we talked about the fact that like wrappers, when I say a wrapper, meaning that you take an AI tool and you wrap something around it.

Gavin: Oftentimes it's not the most valuable thing, but if you do something that you can do a lot of things in one place. And you provide insane value for people actually think it's pretty useful.

Kevin: [00:38:00] We could spend 20 minutes telling you how to manually do this and the results would be questionable and you would have a lot of technological pitfalls along the way.

Kevin: You could just go and support Javi and use the tool and it's going to work. And one of the really, really impactful things, Gavin, that I saw was architectural drawings where you could just kind of sketch an outline of something and then use style transfer to fill in the drawing. The shrubbery and the tiles really bring it to life.

Kevin: , I'm seeing people make basic 3d models of scenes with just like cubes and spheres and whatever else. And then saying, make this look like a dragon soaring through the skies. , so Magnifi AI is the site. You can go over there, sign up, use our promo code. I'm just

Gavin: kidding. Do we ha Oh yeah. We should get a promo card. Okay. It's not an ad. Not an ad. Uh, do you know what time it is, Kevin? Sure. You know what, you know what week it is?

Gavin: It is

Kevin: Oh God. It is

Gavin: this week. And robots. This week in robots! 

Kevin: If you back us on Patreon at the 25 tier, you can punch

Gavin: That's right, exactly.

Gavin: We got two robot videos this week. We're moving through these quick,

Kevin: [00:39:00] We are. By design.

Gavin: Talk to me about the, , Espresso Robot that we're looking at

Kevin: great. Let's talk about the Espresso robot. , listen, NVIDIA. They want to power robots. One of the things that Gavin just said was that, uh, robots can learn through tele operation.

Kevin: Well, this was an impressive demo of a robot using an off the shelf.

Gavin: coffee maker.

Kevin: I think like a little Bambino Espresso?

Gavin: Yeah, sure, the little Bambino.

Kevin: grinds up top and you put the little thing in and it comes down to the puck and then you use the milk wand to make the froth and the whatever else it is like a standard little machine

Gavin: a terrible word. Milk wand. By the way, I think that we should change that.

Gavin: Keep going. Keep going. I think we need to brand it. Keep going, keep going.

Kevin: need AI for Humans branded milk

Gavin: Milk wand. Okay, keep going. So

Kevin: if you're seeing the video version, you've got two robotic arms with little pincers and they're going to go out and look, it's a delicate operation. It's a tiny machine, but someone is tele operating it and it's moving at a speed that I would deem acceptable.

Kevin: , I wouldn't be excited to be making my coffee at this speed, but I wouldn't be disappointed letting a [00:40:00] robot do it for me at this speed. And so this is tele operation, very delicate, fine motor skills, and it's working and it's doing it. And again If you can now train machines by doing this, well, man, you could just throw on some goggles and do your daily routine and maybe the robot arms in your apartment are gonna know how to put things away the way you like or how to make your coffee the way that you want it.

Gavin: again, the thing I, whenever I see this and as you know, what is the thing I hate most to do in my home?

Kevin: What?

Gavin: I think you would remember if

Kevin: it. Oh, uh, drain the milk wand.

Gavin: No, that is not the thing I hate to do the most is to do my socks. What I would love is a robot to do the socks and put the socks together. And I guess if I trained it enough to do the socks, it could do it. Draining the milk wand is something I never want to hear saying again.

Gavin: I don't want to hear you say that ever

Kevin: you over modulated when you said that. That's double

Gavin: I did.

Kevin: Yes, you're doing it again!

Gavin: Oh, what? Oh no. Well, hopefully I haven't been over modulating this whole time.

Kevin: Only when you say, drain the milk one. Okay, what? I'm bleeping all that. Then, uh,[00:41:00] 

Gavin: going to bleep it?

Kevin: You think we should say, drain the milk one that many times? No, probably not. Yeah,

Gavin: Fair enough. All right.

Gavin: The last video we have is a very cool video from figure one, which is a well known now robotics company. It came out of stealth not that long ago, but they are doing a giantly cool robot and what they've done is integrated chat GPT in their robot. And can I tell you, Kevin, one thing that makes chat GPT feel a little scarier is seeing it in a

Kevin: a robot form.

Kevin: Yeah.

Gavin: because this robot, when you watch it, hold on,

Kevin: And like, I know some people were more prescient than us. Like, they were concerned about chat GPT to begin with. But yeah, when chat GPT has like a hand that could throttle you by the neck. Suddenly , you're glad that we're aligning these things a certain way.

Kevin: So when you say, Oh, they put chat GPT in a robot, you think you're just having a casual conversation with the robot and what they're doing here is using GPT vision to analyze the scene in front of them. And then the robot is acting on instructions given to it by the [00:42:00] GPT.

Gavin: exactly. And there's a man standing across from the robot and he's sitting, the robot sitting in front of a bunch of dishes, and the man says, please clear the dishes and put this specific dish away.

Gavin: You can see in the robot's face or whatever, the screen, the chat GPT interface is the same that you would see on the phone and it's listening. It listens. It takes it in and then it responds and does the action. So what's cool here is,

Kevin: hands them an apple at

Gavin: yeah, yeah, you're using chat GPT to direct a robot and it's basically operating as the brain or the, I mean, you can call it the consciousness or whatever you want to call it, the direct, the robot is being directed by chat GPT.

Kevin: Yeah, it sorts dishes, it hands the person an apple, and what's really crazy is that if you watch it at, like, point three speed and zoom in, you see a flash frame of red across the robot's face when it's thinking about tearing the human apart and wearing its skin as a

Gavin: not drain milk wand.

Kevin: And that's This Week in Robots.

Gavin: Reminder before we go [00:43:00] away,

Kevin: we gotta tell people to like, to

Gavin: like

Kevin: leave a comment.

Kevin: A five star review. If this is your first episode, Bye. Bye. So

Gavin: Yeah, if it's your first

Kevin: Percept episode? Watched? He was so saucy today. I just

Gavin: a couple other ones.

Kevin: Man's on the Zinn. Me. He was on the Zen. You struck me.

Gavin: Oh, come

Kevin: Oh, okay. You we

Gavin: gonna do this? Are we

Kevin: Are we, we're gonna do

Gavin: this? All right.

Gavin: Bring up Brenda, tell Brenda to, to ask Kevin why he's being such a wimp. That's what I want

Kevin: to know. Okay. And the toxic masculinity is coming out of his pores.

Gavin: Tell him what, see what she thinks

Kevin: about, you all hear.

Gavin: See what Andrew

Kevin: I'm going to reload Brenda. Um, but sincerely, this, this podcast is a, is a wild one because we're here.

Kevin: We're delighted to still be able to do it. 

Kevin: We tried our best, believe it or not, this was the best we could do with the room and the tech that we had. Uh, we both flew from LA to make this happen. So, uh, please. Like, subscribe, engage, leave a comment, five star review on Spotify, five star review on Apple Podcasts.

Kevin: If you leave one there with a comment, we will read it in this part of the show. Um, tell your friends and your family, we have a newsletter that I have not [00:44:00] written this week, but you can sign up for that. The link is over on AIforHumans. show. That's our official

Gavin: By the way, if anybody wants us to speak, now we are speakers.

Gavin: We are professional speakers, Kevin. So if you have somebody that wants to speak, we can speak.

Kevin: Bye, everyone. I'm done. I'm done. Goodbye. Bye, Gavin. Thanks, everyone. Goodbye. Goodbye, everybody. Goodbye.