← AI For Humans: Weekly AI News, Tools & Trends
Loading summary
Transcript93 lines
- [00:00]
A
Nvidia is flexing its muscles at its annual GGTC conference. As CEO Jensen Huang says they will do over a trillion trillion dollars of business.
- [00:09]
B
Oh, do two pinkies to the mouth. We're gonna walk you through Jensen's plan for total domination. From new robots to open claw integration to something that has to do with cake.
- [00:21]
C
This conference is going to cover every single layer of the five layer cake of artificial intelligence. From land, power and shell, the infrastructure to chips to the platforms, the models.
- [00:34]
A
Yum yum, Kevin. We'll also discuss what this means for you, the human. When everyone has an agent and all GPUs are going towards inference, demand just
- [00:43]
C
keeps on going up. There's a reason for that. Finally, AI is able to do productive work and therefore the inflection point of inference has arrived.
- [00:56]
B
And gamers, don't worry if you're mad that you can't ever new graphics card ever again, prepare to be big mad. Because AI enhanced graphics are here whether you want them or not. That's right. DLSS 5 is here. And Gavin thinks they're super cool. And all of you are wrong. He said that to me on the phone.
- [01:14]
A
I did not say that. I do think they're interesting, Kevin, but I did not say that.
- [01:19]
B
He actually did say that. In fact, he said he wants to use DLSS on us. No, no, I didn't.
- [01:23]
A
I didn't say that.
- [01:24]
B
Oh, this is actually, I'm. I kind of. I kind of dig it.
- [01:30]
A
This is AI for humans, everybody.
- [01:32]
B
Is it?
- [01:38]
A
Welcome, everybody, to AI for Humans, your weekly guide. Actually, your bi weekly guide to the world of AI. Yeah, what do you know? We're doing. Hey, we're going by Kevin.
- [01:46]
B
We're schedule fluid. Guys, you were correct. Comment section. You got us.
- [01:52]
A
Today. We are talking about Nvidia's gtc, which is their big event. Kevin, did you know that GTC D stands for GPU Technology Conference. Did you know that? I did not know that.
- [02:02]
B
I didn't know it until about five seconds ago, Gavin. And I wish I could use that Adam Sandler remote and rewind and unlearn it.
- [02:10]
A
The click. You're talking about the click remote from that movie.
- [02:13]
B
Click Black would have been the easier draw. Sure. But I want to rewind. TiVo.
- [02:18]
A
Doesn't matter. We're going to talk.
- [02:19]
B
Yes.
- [02:19]
A
Okay, let's talk. Let's talk about gtc. I do want to jump into this. So first of all, Jensen comes out in the leather jacket and. And he is just boss manning this conference. He comes out, he's doing Everything. You know what I found out? He doesn't even use a prompter for this thing. So he comes out, he's got all this memorized. But I want to play one little clip here before we start to just kind of get a sense of what he's going for.
- [02:40]
C
Well, I'm here to tell you that right now, where I stand, a few short months after gtc, dc, one year after last gtc, right here where I stand.
- [02:56]
B
Ptsd. No, it's. Keep going, keep going, keep going. It was PT. It was Pacific time.
- [03:05]
C
I see through 20.
- [03:08]
A
Keep going. Keep playing it. Keep playing.
- [03:10]
B
Is he going to say anything other than acronyms and individual letters?
- [03:13]
C
I'm dying 27. At least $1 trillion.
- [03:21]
A
$1 trillion, Kevin. $1 trillion of cash.
- [03:25]
B
Is that what he's. Cash money for new jackets? What is the 1 trillion Nvidia's made? $1 trillion.
- [03:31]
A
That is the money that is going to go through Nvidia, which is crazy to me. And it just. There was a slide they put up where they just show how much of what they're in. Like, they are in every single factor of the modern tech economy. They're in every single factor of future tech economy. Demand is just going. It's like insane to me when you watch this.
- [03:52]
B
To be clear, for the audio only. Automotive, financial services, health care and life sciences, industrial. That means, you know, water, power, you name it, media and entertainment, quantum retail, consumer packaged goods, robotics, telcos. They're in literally everything. And as AI infuses its way into everything, that will include your toothbrush, that will include your dog's collar, that's going to include everything.
- [04:15]
A
And also, Kevin, you know what we'll really include, which was kind of one of the bigger announcements of this conference was DLSS 5. It's going to be in your video games in a much bigger way than you might have expected. Of course, Nvidia comes from video games. He did a little bit of a history thing where he talked to. It's the 20th year of CUDA for those out there who celebrate. So congratulations. CUDA can almost drink a beer. But that is based on video games. We used to make graphics cards just for video games. And guess what, Kevin? They introduced a new thing that is very interesting. It is an up resin tool for video games. It is called DLSS5. And this is essentially like you and I have seen over the course of the last couple of years. You could see up res your video or up res your photo with AI. We'll be showing this on screen. But if you're not watching this, you know, it's a slider back and forth and you see essentially an AI uprest image. When you're looking at these video, this video, like, what are you feeling?
- [05:09]
B
The. The rage of the Internet. I can feel them sharpening their pitchforks.
- [05:13]
A
Take away that, Take away that. What are you feeling?
- [05:17]
B
I think it's super interesting. Look there. There are certainly times where UPRES techniques can breathe new life into something. This is slightly different than previous takes at dlss. If you've seen AI up res graphics in the past, they were just trying to imagine the pixels in between the pixels, if you will, or sharpen out the edges or something like that. This new approach actually takes like the frame color and the motion data and uses AI to add, add, not just enhance, but add accurate lighting and materials and fine detail. And they will claim it's using the original models and colors and textures of the core game itself. So they're remaining true to the artist's vision. This is where some of the division is because when you look at the examples, it looks like it's painting details that don't exist in the core models as they are on the screen now. My parents love Emotion Smooth Gavin.
- [06:09]
A
The they. Oh, they do. Oh my God. I know.
- [06:11]
B
To take a, you know, the old 4x3 video and up res it to 8k.
- [06:15]
A
They're watching Mission Impossible in 4 by 3 in black and white.
- [06:19]
B
All the characters kind of just slide across everything. Everything looks like a soap opera. They love that stuff. And I do think with certain games there's a time and a place for this. There's gamers that will line up for a new texture pack of a re release of a Final Fantasy or whatever else. So I do think there's going to be clearly an audience for this. We've seen already in the very short time that this was announced. There is definitely a hater contingent for this. Right. And I think if you're like a hardcore gamer, just the fact that it adds and it's only like 18 milliseconds of latency, but the fact that it is adding latency makes this like a deal breaker for some people right off the rip.
- [06:58]
A
Yeah. And I will say it's like, you know, kind of the same thing we say with all AI stuff. It's like the worst it will ever be right now. And I think that one of the things I think they kind of screwed themselves over by is like there's actual changes. Like to your point in the background, one of the shots of the Resident Evil shot, there's a sign that says cigarettes and the cigarette sign is gone. So maybe it's just that AI is trying to clean up our lives and not get us to smoke. Starfield, which is a video game that I played. Did you play Starfield? I can't remember if you played Starfield.
- [07:22]
B
I did, yeah.
- [07:23]
A
Yeah. So one of the problems with Starfield is they had these very dead looking characters. Right. When you look at these characters straight on, they were very dead looking. And I will say, Kevin, AI did not really fix that. It made their skin nicer, it made the backgrounds nicer, it did make the environments really cool. There was one environment, Starfield, really beautiful, but the eyes are still kind of dead. And so it's not going to just fix things that are necessarily wrong with it. Digital Foundry is a very great YouTube channel you should be following. They got an inside look at this. Their video has a lot of other stuff. One of the cooler parts that they showed was a shot from Assassin's Creed Shadows, which you haven't played that. That's the one where you're in Japan and one of the characters is up on top of a mount, on top of a building and he's seeing this forest and the forest looks amazing. Like when you look away, like at the, at the landscapes, I think facial stuff is going to be a little bit different. But Kevin, I did want to point out, to your point about the idea of how negative people are, those comments are watching people just like light the Internet on fire. In fact, one person, I want to call this one out, Fampas F A M P I S said quote Apply the slop filter, please. Quote said no one ever. So you are going to see a lot of negativity.
- [08:30]
B
I think we use the slop filter on our webcams. Every recording we apply it directly to the forehead. Listen, we. When we talk about. And we. You do say it often and I think it bears repeating that this is the worst that it will get right. It's easy to look at those examples and go, oh, there's a halo of light around the character as it moves. Oh, the, the background signage changes subtly. Oh, the, the. The. The rain details aren't exactly in line with the. It's the worst that it's going to be. Yeah, it is going to get better. It's wild to me still that this system understands like hair, skin, fabric, foliage, has understanding of all that and can do it with. I think it was like an average 16 millisecond hit that's going to drop down even more. And I think beyond the, oh, we're just going to enhance our existing game. I think we're going to see games designing for this where they're going to ship ultra low res geometry, ultra low res models and be like, hey, this card has been trained on this blocky thing and suddenly like I darb or the old NBA Jam looks.
- [09:34]
A
Oh yeah, that would be so interesting to think about. Like how low poly could you make something and actually have it come out? Yeah, how low could you go? Kevin, I do want to ask like if, if they didn't have the word AI attached to this, would gamers be excited? Because that's my big question. When you just look at what this is, Sure. A gamer who didn't think this was AI might be like, I kind of changed it. But wow, I can see like so much better. It feels like a whole new gaming system. That's the thing I think is weird to me.
- [10:01]
B
Graphics card have hyperpixels. It got the hyperpixel dragon on it.
- [10:06]
A
Hyperpixel dragons and AI carbon fiber scales on it.
- [10:10]
B
I'd be like, oh Yeah, I need 10, 10 carbon dragons. But no, I think you're right. I think AI kind of automatically yucks a lot of yums. Again, I understand people that are like that's not the artist's intended vision or that's, that's. It's. It's adding things to the game that shouldn't be there. But also sometimes the artists push it as far as they can and maybe they wish that the self shadowing on the eyes of the character would have been more defined. Like we're, we're not necessarily able to say if the game developers themselves are officially supporting the thing, then they're welcome to. You're also welcome to hate it too, but you can't.
- [10:45]
A
Oh, I forgot there was one I saw. Kevin, I have to show you. So I guess Nintendo got involved in this and Nintendo decided that they were going to up res Mario. And I don't know if I buy this one necessarily because it seems to me like this feels a little bit off. Like I'm not sure if I buy this one. So anyway, if you're just listening, go to the YouTube and watch it and check out what we've done here.
- [11:08]
B
Why does it turn him into Timothee Chalamet?
- [11:10]
A
I don't know. This is. Maybe that's the next star of the Mario movie.
- [11:14]
B
This is gonna be a great meme. And if anybody in our comments or on our discord Wants to do a DLSS5 on all of the things even outside of video games. We are here for it, please.
- [11:24]
A
Yes, we are. Okay, so very fast. There was a quick mention of Jensen talking about chips in space, which has been this ongoing conversation. If you remember, a couple weeks ago, we talked about. Elon discussed this idea of, like, all of his chips are going to be in space. Nvidia's working on chips in space. They have a way to figure that out. Although he did mention the fact that, like, they're going to have to figure out how to deal with heating and cooling up there. And Elon's idea was always like, oh, we can just use the coolness of space anyway. That's something they're working on. But, Kevin, the bigger thing that came out of this, I think, was this real open claw integration, like, leaning in. Whoa, what's going on? What are you laughing about?
- [12:01]
B
Just the reality of what you and I have chosen to discuss each and every week to go, yeah, yeah, the chips in space. Yeah, we're going to launch these space data centers. We're going to cool it with the vastness of the vacuum that is space. And there's no timeline or partners, but we're going to figure it out.
- [12:15]
A
But here's the really big thing. I mean, the chips in space thing to me, and I'm curious to know what other people out there think. It feels like a pipe dream right now. I know it's not. I know that there's science that can do it, but there is no actual way to beam down the data yet in a significant way. So it does feel like to me, it's like everybody's got to check the chips in space part off of their resume or off of the thing. Like, I've done gypsum space. Do we do a great week? Okay, great.
- [12:41]
C
Yep.
- [12:41]
B
Okay.
- [12:41]
A
Well, you can tell, like, Jensen used, like, he only talked about it for like five minutes. It was a three hour keynote and he spent five minutes. Anyway, the bigger thing here is that they actually had Peter Steinberger, the creator of openclaw, in the audience. And they have worked on a whole new integration of OpenClaw within the Nvidia platform that they're calling Nemo Claw.
- [13:01]
B
And.
- [13:02]
A
And this is a little. Seems a little strange and weird that we're mixing names here, but. So Nemo Claw is based around the name of Nemo Tron, which is one of their other platforms. But the big idea here, Kev, is that they are trying to make Open claw accessible for SaaS businesses, meaning that you could trust your enterprise data with an open claw service and the idea that this could really open the door. And by the way gents, they made a whole video about OpenClaw. They went like off about open cloud. They really loved it. So they're cloud. I think a big part of it for them is they really want to boost open source open cloud. One of the things Jensen said in the presentation was that it is the. It is the biggest open source project of all time, at least according to GitHub. That it took two weeks to beat Linux in terms of how many GitHub stars it has, which is pretty crazy. Anyway, I'm curious as somebody you've spent a lot of time with openclaw and I've actually been talking to, like I mentioned last week, my brother in law who's been doing a lot of openclaw stuff, where do you kind of stand on the idea that like Nvidia could give OpenClaw a big boost and that it could make this kind of enterprise ready?
- [14:06]
B
I did not see this one coming. Yeah, with the open claw of it all, I think a lot of people that especially the early adopters got into it and they're like, okay, yep, this will go. Especially once OpenAI got involved, right. And kind of snatched up Peter, we were like, okay, this thing's going to have legs here and then the bigs will go make their own version. I did not, and this is my short sightedness, I did not count Nvidia as one of the bigs that would take this and run with it. In a way I figured Anthropic is going to put more cloud like features into their suite. Gemini and Google, they're going to do the same. Microsoft have an offering. I didn't expect Nvidia to do it. And that to me is so fascinating because we saw other slides where they announced all the big foundational partners that they work with.
- [14:48]
A
Right.
- [14:48]
B
Gavin, they had the vodka side and they had Codex and they had all the big partners are there but like don't forget they can also serve the open source models and they have 1 trillion running through their veins so they can spin something like this up. And I think, look, making it enterprise secure in addition to the stability issues and everything else, but I think making it enterprise secure was sort of the number one thing because that's what held even some normies back from using it was this notion of like I don't know, that I can trust it with the keys to my personal kingdom.
- [15:19]
A
Yeah. And what I find interesting about Nvidia is they win no matter who wins, right? Like that's the thing you can take away from a presentation like this. Like they are winning across the board and all they need is one of these people to survive. Like to be. And by the way, if in an open source world, Nvidia does great because as Jensen pointed out again and again, inference runs on GPUs, right? Meaning that inference, if you're new to someone, new to the space, that's where the AI has to think, think about things before it gives you an answer. It's not just the training of the model, it's the real time calling of the data going back and forth. So for Jensen, Open cloud is great, right? Because it brings more people on board and it brings more open source models on board. And it doesn't matter how much they're paying for the OpenAI token or the cloud code token. It's more about getting more and more people on board with the AI train.
- [16:06]
B
Well, here's the other thing. You go to an Apple store and they're sold out of Mac minis because people were snatching those up to just run Open Claw, which it's crazy overpowered for that, but they did it regardless. Nemo Claw gonna run on RTX graphics cards, gonna run their RTX Pro, gonna run on the DGX station, probably gonna run on their, their Spark, their little supercomputers. So they have a vested interest in making sure it runs really, really well on their hardware. Because I bet they saw that moment and. Oh no. What's going on?
- [16:33]
A
I'm just thinking, Kevin, not financial advice, but am I supposed to buy Nvidia stock at this point? Is that what I'm supposed to do here? It's got a flag.
- [16:40]
B
Space chips, the space Chance, you know,
- [16:44]
A
you should buy is a little bit of investment in this YouTube page by liking and subscribing right now. We've got some more to talk about in a second, but just take a moment. Click that like and subscribe button. We are going to come to you twice a week now. That is our goal for the next month. Yes, you've committed to it, Kevin.
- [16:57]
B
No, for a month. No, it's a month. I just realized someone's going to do the DLSS5 on me and it's halfway through. And then the other half of my face is Kevin Rose.
- [17:05]
A
Oh, man. Well, we'll see what happens. Maybe I can be Alex Albrecht. Maybe there. It's like all the old people unite in one big thing. We could all be together and One big. One big retirement home. You and me will be doing our podcast at one side, and Kevin and Alex will be doing it the other side of the retirement home and be
- [17:20]
B
like, gavin is Alex Albrecht.
- [17:23]
A
He's gonna love that.
- [17:24]
B
But, yes, sincerely, please support us. Leave a comment Thumbs up. Subscribe and if you want, we got a Patreon and a Buy me a coffee and a whole bunch of other stuff. And subscribe to the newsletter.
- [17:34]
A
And we got a great thing coming on Friday, so don't forget, we have a second show coming on Friday. All right, Kev. The other big thing that came out of this in GTC was they brought out the robots. And the robots are always really interesting. I remember you and I covered this last year. Maybe it was last year, the year before, they talked a lot about simulated environments for robots. Right. Remember, there was a whole thing about how you put a robot in a simulated environment. Well, this year it's clear that a lot of these robots are actually doing this. And the most interesting part of this to me was, you know, the little Olaf robot that they've rolled out across Disneyland. Now it's, like, about. I don't know, about this tall, like, up on my desk. Like, this tall. They had the Olaf robot come out. But the more interesting part was looking at what they designed for the Olaf robot before they put it into practice in Disneyland. They had a whole simulated environment where it was able to kind of learn how to walk through kind of dirty snow in some way and different things. And, like, that feels like the future of this kind of integrated robotics AI. And right now, Olaf doesn't feel like the thing that can kill us, you know, because he's cute and he's got those little sticks for arms. But at some point, someone's going to be simulating that in a much bigger way in a practice of war. But we're seeing just the cute version of it right now, Kevin. It's just a very cute version. So I appreciate it.
- [18:42]
B
You know, there's a workstation somewhere where they replace those little twig arms with AKs. They've got tiny little Olaf learning to waddle over marbles with two death machines strapped to his arm. I love it.
- [18:53]
A
Let's do it.
- [18:54]
B
You think they're going to replace the coal buttons with thermite and Olaf just sort of waddle in?
- [19:00]
A
Let if somebody in our audience try to make the Terminator shot of where the Terminator steps on the skull, but make it with Olaf, and we will give you a special shout out. We'll give you a special shout out.
- [19:11]
B
Yeah. I want to see when he comes down. Like when the little bubble forms and he's teleported. And it's usually like a naked Arnold. Let's see Olaf.
- [19:17]
A
Naked Olaf. You want to see naked Olaf? No. You want to see naked Olaf?
- [19:22]
B
I don't know what's under the snow. I guess he'd be melted at that point. The point is I want to see
- [19:26]
A
what is under his snow. I guess it's just snow.
- [19:28]
B
Nothing. It's just snow.
- [19:30]
A
Are we sure? Are we sure there's nothing under the snow?
- [19:33]
B
Well, the robot, they're definitely mostly falafel and. And Legos.
- [19:37]
A
Yes. The other thing that's very cool about gtc, there was a video game that came out from the same guy that made There was a great minesweeper thing about the strings of horror news that came out about a week ago. And he made that videos like a casual game. He made a video game called GTC which is good till canceled. And this is essentially a Vampire Survivors game where you play as Jensen and you fling chips at people coming towards you and investors looking for money. I played this, Kevin, for longer than I would like to admit. It is very fun. It is. It is a. It is a web based game and it goes on. It's not like you can level up. There's level ups and power ups throughout it. So, like just a very fun, cool vibe coated thing. And the last thing I want to say is there's a great podcast you should all listen to to give you some more insight into this. This one. This is this podcast too. But Dwarkesh Patel did an interview with Dylan Patel, who Jensen shouted out in the keynote. Dylan Patel runs Semi Analysis and Semi Analysis kind of does a lot of data and research about the chip world and the, and the energy world and all the stuff that kind of matters in the hardware world. And in this interview, Dylan talks a lot about the bottlenecks may not be energy as much as people think energy is going to be. Like, there's a lot of conversations around like 1 gigawatt versus 5 gigawatts nuclear reactors.
- [20:48]
B
We need space labs. Okay. Yeah.
- [20:51]
A
He was saying the bigger bottleneck is more likely going to be the small pieces that have to come together to get these chips. Like there's a. There's a company, I think, in the Netherlands that makes just the small piece of all of these kind of processes that have to come together to make the fabs, which is the machine that makes the chip. All of this stuff is this crazy big thing. And so there's all these little pieces that could go wrong. And so just to be clear, like ram, by the way, or graphics cards too, right? Like all this stuff. But just like, as a little bit of like, you know, Nvidia and Jensen is here kind of selling the big sell. That's a nice little kind of way to kind of look at some of the realistic situation we may be in in the next two to five years when it comes to all that.
- [21:36]
B
I think it's a bubble. What are your thoughts? Fab. Are you okay?
- [21:40]
A
But did you just guess? Was that the bubble? The gas you just.
- [21:43]
B
That was me just being. Just being real.
- [21:46]
A
We'll see you all on Friday. We have a very special thing that Kevin has been vibe coding that we're going to. We're going to premiere on Friday and maybe have playable at some point. Right? That's the idea.
- [21:54]
B
Just really. I'm going to release it open source so anybody can do it. If you ever want to, like, play a game of Doom and then just whisper new enemies and weapons into the game, like, I don't know, shoot a juicy hamburger at Guy Fieri.
- [22:08]
A
Okay, I can't wait. Also, like, and subscribe this video, we are going to come to you twice a week. We're trying to make a little shorter video. This will still go up on audio, but like and subscribe. We're very happy you guys are here with us. Thank you so much for watching and we'll see you on Friday. Bye Bye.