The AI Sandbox - Testing Cool AI Stuff Live

Sam. Sa. We're going live. Welcome to the AI Sandbox. We're going live. Going live. Welcome to the AI Sandbox. We're going live. Welcome to the AI sandbox. We're going live. Welcome to the A.I. sandbox. We're going live. Hosted by Matt Wal. We're going live. Hosted by Mad Wolf.

Yes. Yes. I swear that does not get old. And we are indeed going live. So I'm all hyped up. That music always bumps me up right before I go live. We are actually doing something a little bit different today. This one's going to be a fun stream, testing a new flow, testing a new software, not actually using obs this time around. I'm using a tool with some other stuff and you'll find out why here in a moment. But the idea behind this show is I want it to be even more interactive than normal. I want to be able to pull up your comments on screen. I want to be able to interact with you guys. I want to be able to have a little bit of back and forth communication. So on this live stream and probably future live streams, I'm going to have a actual live producer, slash, let's call him, color commentator on the shows with me. And him and I are going to kind of bounce ideas around off each other. We're still going to do the shows just like we've been doing them, where we're going to talk about the news, we're going to play around with some tools, we're going to respond to comments, answer questions, things like that. But now I've got like a sort of Jamie to my Joe Rogan, right? I've got somebody that can pop in and, and help me, you know, talk about some things. When I run out of ideas, he'll come in and say a few words and you know, he's managing the cameras and things like that. So he'll be, you know, switching between my screen and my camera and other videos and things like that and should make it a lot more fun and interactive. He's, he's one of my best friends in the world. His name is Randy, but his friends and the online community call him Buffalo. And he's going to be joining me a little bit on this stream today and helping keep things fun and lively. I don't know if he wants to pop on the screen and say hello or hey, it's the Buffalo. Hello.

First off, dude, that opening, I'm holding drumsticks, dude. I was like, I was in it, man. I love the opening. I love the energy Your Monday streams are just going to be so much fun to be here with you. I've been in the AI space for a day, and then I leave for a year, and then I come back for a week. And so every time I leave, I go to your. Your YouTube channel to listen to what you're covering, man. And every time my brain just explodes. So I'm. I'm going to be learning a lot alongside all y' all in the comment section and hope to help up upgrade this production value of your stream. Bro, this is gonna be a lot of fun.

Yeah, absolutely. So Buff, as I like to call him because I'm too lazy to say the full buffalo. That's like way too many syllables. You know, he's. He's got a production background. He actually does, like, camera work for the Mariners and the Seahawks and the Kraken and worked with doing production at Microsoft. So, like, I've actually got a real pro in the mix for once instead of me, just like, what does this button do? And then, like, trying to figure it out and then go, I really shouldn't have pressed that button. He actually knows what he's doing. So it'll be fun to have a little bit of different production value. Also a little bit of a. A second voice to bounce things off with from time to time. So should be real fun. And I appreciate you, Buff. Thanks for, for hanging out and doing your thing.

Yeah, you're very welcome, sir.

A lot of fun.

I'll let you take it away.

Cool, man. So let's go ahead and do some screen share and we'll start with. With the way we start in the past and, and talk about a little bit of AI news that's been circulating. There hasn't been a ton of news over the weekend. I'm not expecting a ton of news to come out today. Obviously, today was the presidential inauguration, and pretty much the entire tech world was there. I don't know if you guys watched any of the inauguration video, but, like, you know, Tim Cook was there, Sergey Brin was there. Sundar. Sundar was there. From Google, the. Mark Zuckerberg was there. Elon Musk was there. Joe Rogan was there because, you know, he's in politics now, I guess. But yeah, so, like, I don't think we're going to get any tech news because they're all there. Also, today is Martin Luther King Day, so happy Martin Luther King Day to everybody. Which I know also means a lot of kids have today off school. So anybody that has kids, they're not you know, sharing news or building things today, I guess. I don't know. I just think today is going to be a slow news day because of everything that's going on in, in the country right now. But there are a few things that I think will be fun to talk about and then we'll spend a lot of time really, really playing around with some tools. There's a few tools I've been meaning to dive deeper into and just sort of play around with them and we'll mess with them on stream and see, you know, see what we can make them do. I'll take suggestions from you guys and, you know, have, have some fun with this stuff. But the, the one kind of newsy sort of thing that came out over the weekend that a lot of people have been buzzing around was this article that came out from Axios called Behind the curtain coming soon. PhD level super intelligence. This basically, I feel like this is sort of like an overhyped news story personally, but basically it's saying that that OpenAI is going to be announcing a next level breakthrough that unleashes PhD level super agents to do complex human tasks. And I don't know, it's interesting to me because this article came out right? And then like literally this morning, Sam Altman tweets, twitter hype is out of control again. We're not going to deploy AGI next month, nor have we built it. We have some very cool stuff for you, but please chill and cut your expectations 100x. And it's kind of interesting that he's the one tweeting this because I think anybody that follows AI really, really closely sort of feels like Sam Altman is the hype machine that is creating all the hype. You know, so there's, there's been a lot of, a lot of people commenting about the fact that, Sam, you're the one creating all this hype. And there's an article on Axios right now basically claiming you've got PhD level super agents coming soon. Like, why are you telling everybody to chill out when you're also the one sort of creating all of the hysteria over AI? I mean, we can. I don't know if I'm going to like actually dig through all of his old tweets and stuff, but I mean, like, here's one prediction. The OH3 arc will be something like, oh, damn, it's smarter than me. This changes everything. Ah, so what's for dinner? Can you believe how bad oh3 is? I mean, to me, he's like, this is already him like, like hyping up what's coming next and then he's going and making a tweet saying everybody chill out. It's like you're the one creating the lack of chill. I don't know. I'm curious what your guys thoughts are on this. It's just been kind of interesting to follow. That's really the only like really interesting news that came out over the weekend, in my opinion. I think, I think at the end of the day essentially what, what this news is all about is that he's scheduled a closed door briefing with the US government to basically show what he's been working on. And for whatever reason Axios is like hyping this up into like, oh, he's going to show off AGI to them. But that's not the impression that I'm actually getting out of all of this. Let's see. I'm taking a peek at the comments here real quick. Do you think the 4o for task is actually useful? I do. I think we'll play around with it a little bit. I'll show you how I've actually played with it. I think that's pretty useful. But we'll, we'll get to that in a little bit. I want to, I'm going to stick on news for a little bit first and then I'm doing my best. And this is one of the reasons Randy's going to help me, Buff's going to help me. Here is like, I'm doing my best to try to keep these live streams a little bit more segmented. Like maybe just talk about news for the first 20 minutes and then go into playing with tools for a bit and then go into like, all right, let's just. Wherever the chat wants to go, this is where we just let the madness happen and unload craziness and talk about wherever we want to go. In fact, Buff and I were doing a test stream on Friday and we accidentally went live for like what, five or 10 minutes. And during that test stream I think we had 40 people show up by accident while we were just testing things. And I was literally drawing stuff in Microsoft Paint during that test stream just to test things. And I feel like we can equally go off the rails, but we'll kind of do that like in the final hour. Probably see just scrolling comments here. I want, I'm curious what you guys think about all this. Oh this, all of this hype and then Sam telling us not to be hyped. That's not like Sam, he's a Cheerleader. Let's see. I think there's heaps going on behind the scenes. Yeah, totally agree. I think there is probably a lot going on behind the scenes. Cool. All right, I'm going to go ahead and move away from some of this stuff here. I'm going to actually check into my future tools website, pull up the news, see if there's any other interesting news that kind of came out. As you know, if you watch my Friday news videos, what I typically do is record those on Thursdays, put them out on Friday. So any news that came out on Friday usually get missed. Let's see here. So anything that came out on Friday, I'm going to see if there's anything worth talking about. OpenAI enhances ChatGPT with simplified user controls. So it looks like we're getting an update inside of CHAT GPT. This is like super, super minor update. What traits should ChatGPT have? You can tell it to be chatty, witty, opinionated, straight shooting, Gen Z, poetic stuff like that. That was something that OpenAI announced on Friday. I mean, cool. Like I, I don't know how, how exciting that is. I don't really, I don't really need my chatbot to be sassy to me but you know, if you do. Cool. Let's see. This one was interesting on X. I don't know if you guys saw this over on X. This guy actually used Claude build a neutron producing.

He's gone. Hopefully he shall return. Until the meantime, please stand by.

Sam.

Well, that's embarrassing.

I'm glad I had that little technical difficulties built. Yeah, I was like, oh, I thought it was my computer until I realized.

I was the only one left. No, no, my. Apparently I had a little technical difficulty and my Internet blipped. That's the first time that's happened in a while. But the good thing is you were hanging out of the stream. So when my Internet kicked me off, at least the stream didn't completely shut down. So yeah, one of the, one of the benefits of having somebody else help manage the stream. All right, so I'll get back to. I was talking about, of the. I don't know, I don't know where it cut out unfortunately. See, how do I get to my screen again? How do I do this? Sorry guys, I don't know what I'm doing here. All right, so yeah, this is, this was what I was showing right before my Internet blipped out on me for a second. This guy here posted on X that he built a neutron producing nuclear fuser in his kitchen using Claude. So he actually built this nuclear fuser in his kitchen and he did the whole thing by essentially putting all sorts of information about how to build nuclear fusers inside of the project knowledge and then asked it questions until he was able to build it. To me, that's pretty wild. Also pretty kind of scary. There's something a little bit scary about that to me that you can just use Claude to build this kind of stuff. But pretty, I mean, pretty crazy that somebody was able to do this. Blink 4 times if FBI AI robots are in your room. No, I'm good, I'm good. Don't worry. Yeah, somebody's asking, what does it do? To be honest, I'm not really sure. I think it was more of like this proof of concept. Somebody down here said, see, where's the. Somebody asked the question, like, how much energy does it produce? And he basically said next to nothing. He was just trying to like prove the concept. So yeah, I don't know. Let's see here. OpenAI's O3 mini model set to launch. Okay, so Sam Altman did basically say that O3 is coming soon. He said thank you. For the external safety researchers who tested O3 mini, we have now finalized a version and are beginning the release process, planning to ship in a couple of weeks. Also we heard the feedback will launch API and Chat GPT at the same time. Mitch wants to know, will it wash his dishes? Probably not, probably not. So, but basically Sam Altman's telling us that within the next couple weeks we'll get O3 mini. Sam Altman lately, I feel like, has a tendency to over promise on things like this under deliver a little bit. Like if you remember when he said advanced voice mode was coming in a few weeks and then it took like seven months for us to get it. So whether or not we actually get an O3 mini or not, I don't know. But somebody did ask, are you able to speak on how capable O3 mini is compared to O1 Pro? Very specifically 01 Pro, not 01. So 01 Pro is the one that you gotta pay $200 a month to get inside of chat GPT. And he said that O3 mini is worse than 01 Pro at most things, but it's gonna be a lot faster. So if you have Zero1 Pro, if you have the $200 a month ChatGPT plan and you have access to this Zero1 Pro mode, you're actually already getting a model that's better than this O3 mini that they say is coming out in the next couple weeks. I think more excitingly is actually the launch of the API than anything, because when they launch that API, we're going to start to get 03 and probably 01 inside of tools like, like Cursor, like Windsurf, like, you know, all the various development tools that leverage AI once an API is available. Because O1 is actually pretty dang good at coding. In fact, Nathan and I did a recent podcast episode on the Next Wave where he was showing how he's basically been building a really, really cool looking game using 01. So it's really good with code, but you kind of have to still copy and paste from Chat GPT. It's not integrated directly into Cloud or Windsurf or any of those tools yet. So once we get the API, that to me will actually be sort of a bigger news than actually getting access to O3 mini. Especially since the O1 Pro that's already available is probably better. Let's see, I'm trying to figure out the best way to kind of keep up with these comments. The problem with this new tool that I'm using is the comments are like the furthest side of my screen, so they're kind of hard for me to read. But I'm trying to try to skim through and see if there's anything you guys are saying. AI is currently searching for new semiconductors, materials, metals. Yeah, I don't know if you guys saw this. I talked about it, my last week's news video. But this matter gen is really, really interesting as well, where it's basically like a, an AI tool, like a stable diffusion or ChatGPT, but you basically prompt it to generate new materials for you. So that's pretty wild that that exists. Let's see, I'm wondering, Buff, if maybe any sort of like questions that pop up we could just toss into like the starred folder or something like that so I can, I can see them easier because I'm really, really struggling to skim the comments over on this, this app.

Yeah, I've got some starred there for you. I just didn't know if you wanted to bounce some different subjects.

Yeah, yeah, yeah. I just want to make sure that like when I'm on a specific like piece of news or something like that, if somebody is actually commenting on it, I can, I can sort of pull it up and comment on their commentary. Let's see. Andreas is asking, will you dive deeper into NSFW nerve gear virtual world? Probably not. I don't do anything NSFW on my Channel try to avoid it. This. I. I get a lot of people when I go to conferences and things like that that tell me like they watch this channel with their family, their kids, you know, watch my videos with me and stuff like that. So I purposely avoid any sort of, like, not safe for work generators and AI girlfriends and all of that kind of stuff. I know they're out there. I'm sure there's plenty of other channels that will talk about them, but it's just one of the topics I choose to. To not dive down. Not. Not one of the areas of AI that I'm actually that excited about. Deep Seek just released the R1 model update, and it's amazing. Yeah, I'm not sure what the best place to play with the R1 model is right now is. Do they actually have like a. Is there actually a place to mess with it right now? Do they put it up on hugging face or something like that? Are there any hugging face spaces? I don't know what Janus Flow is. These ones have errors on them. I'm actually not sure the best platform right now to mess around with Deep Seek. I did see some people were asking about using Deep Seek, but wasn't quite sure the best place or way to use Deep Seek. All right, so test Deep seek. They say R1 matches 01 test deep seek. There's a lot of bros in this message. Should I just read the whole message word for word? Test Deepseek bro. They're saying R1 matches 01 test deepseek bro. They're saying R1. Okay, it's just repeating over and over and over again. All right, so I'm getting spammed to test R1 deep seek. Basically, I don't know how to test R1 deep seq. Okay, so somebody's saying it's available directly on their website. All right, R1 Deepseek. Let's go ahead and see if we can get to it right from their website. Start now. It's gonna make me log in one. All right, so here's Deep Seek. Is this. How do I know if it's using the newest R1 model? Okay, so deep think use deep seq R1 to solve reasoning problems. So if I select Deep Think, that's using the R1. All right, so somebody give me a prompt to test. One of the things I've constantly talked about is I struggle to test new large language models because I never know what to test to actually see if it's really good or not. I need like a complex problem to Prompt to see how well it does. Can Deep Seek pull government patents asking for a friend? I'm not sure I'm going to test that one on stream. All right, I guess I'll think of something here. Need. I need some prompts to test. Tell it to code malware that'll give it sentience. Yeah, that sounds like a really, really smart idea. I never know what to test with large language models. That's the problem. I use large language models like chat, GPT and cloud all the time, but I use it for like, basic troubleshooting tasks and like, brainstorming and stuff. So when everybody's like, test this new large language model, I'm like, okay, tell me what to test it with. Because the large language models I use every day, like, they work just fine for everything I need it for. All right, write the Snake game using PI Game. Right, Snake using PI Game. So it's actually explaining what it's doing as it's doing it. That's something that 01 doesn't really do very well. The only problem is I don't know how to test if this code works or not. I guess I'll have to open up like versus code or something. And let me see. All right, so it wrote some code here. Let's see. Basic sneak movement using arrow keys. Food. Because Deep Seek is not going to do what like Claude is going to do where it just like will run the Python file directly in my browser for me. So this is like way more in depth that I really want to test here. All right, somebody give me in. Yeah, test JavaScript instead. That's probably a better idea because I don't really want to start building up stuff inside of VS code right now. All right, so let's go ahead and build the snake game in JavaScript. That's going to be a better test because I can actually run JavaScript a lot easier in my browser. Earth is 8 minutes and 20 seconds from the sun. How fast is the earth revolving around the sun? That's another good one. All right, all right, let's test this JavaScript though, because I can actually do this. I think it's probably going to do it all in line for me. So that'll be good. So I'll copy this. I'll go ahead and toss it into a. I'm just going to do this directly in Notepad here. We'll toss this code into Notepad. I'll save this. I'm just going to throw it right on my desktop and we'll do Snake, Snake, HTML and now let's open my desktop. All right, so the. What. What's going on here? I'm not very good at this game, but it seems to be working. Moving too fast. Okay, there we go. Yeah. Work one shot. It worked with one shot. That's pretty cool.

Nice.

Nice. All right, so, yeah, Snake and JavaScript worked on the first try. I'm pretty. That's pretty good. All right, let's go ahead and test this other one. Somebody said. What was it? Something about the sun being so far from the Earth. Where'd it go? Wait, maybe. Maybe you start it. Where's. Where's that one prompt? I can't find it. Again, This is why I don't test large language models live typically. All right, here we go.

Earth.

There a way for me to. How do I copy it? I guess I just have to type it. I can't. This app won't let me copy and paste it, unfortunately. All right, so Earth is 8 minutes, 20 seconds from the.