FTC Investigating OpenAI and Anthropic AI for Felonies - OpenAI is DEAD and Dario is Going to Jail

Eli (0:00)

So you wanna see something pretty cool that you can do with artificial intelligence?

What, you're here just for drama videos? Well, suck it, you're gonna learn a little bit about how this crap works, cause this is what I actually care about. Anyways, I've been working on Model Context Protocol for a class that I want to do for the new education series I will be doing in Baltimore, Maryland. And what's really cool is I've been able to connect MCP server directly to Claude the desktop version. So this is actually installed directly on my desktop. And what's cool here is I have this MCP server Python script, and basically it allows me to interact with my operating system. So basically here, what we're able to do is we're able to pull out top. So basically, how memory and how CPU usage is doing, we can ping something if we want to, and we can also see what the storage is like. And so what's really cool here is I can actually connect my Cloud desktop software directly to that MCP server. So I can actually now ask this application what's going on with my computer. So I can say, how is my CPU doing? Question mark, right? So this isn't on what's going on with Reddit or Twitter or whatever else. Look, it actually wants to use my script. So I'm just going to do allow once. It's going to go through. It's now going to run the top command. It's going to bring the values in from the top command. It's taking a little bit of time because I'm not actually paying for it. This is even still the free version of Claude that I'm using. And see, your CPU is in good shape. It's about 86% idle with 5.6% going to the user processes and 8.6% going to system work. I can say, what is my internet latency? Because I have the thing that will actually test my network connection. Again, I have to allow once, because this is connecting directly to my shell. And no, to be clear, no, I don't trust anybody, including my own coding. Anyways, your latency to google.com is about 18.4 milliseconds, which is good. That's fast enough for video calls, gaming, and general browsing without noticeable lag. So look at that. So you can actually build your own Python scripts that can do basically whatever the hell you want those Python scripts to do. They can connect the REST APIs, they can connect directly to your operating system, they can connect the databases, vector databases, all that kind of stuff. Anything that you can do in a Python script, you can do as an MCP tool. You then connect that MCP server into the Clawed, oh, the Clawed application that you have on your computer, and now you have access to all of that directly on your system, where you can just ask natural language questions. Isn't that amazing? Isn't that cool?

Yes, it is. Yes, it is. But I know, none of you give a damn. You're all sitting there going, that's nice, Eli, oh, oh. You're showing us how the orchestration layer works, Eli. Oh, oh, Eli's talking about how the AI stack actually works. Oh, oh, Eli's talking about how you can actually make valuable things with AI. Oh, oh, Eli, we don't. Yeah, we are here for the drama. What don't you understand?

Fine, we'll get to the drama.

Get all these viewers. They're like, you're just doing this to get clicks. Yeah, that's how you make money off YouTube.

I know the young people don't seem to grasp that question. Taxes or that idea, taxes, mortgages, grocery bills.

Still not sure how the younger generation eats. I'm still at a loss to understand. Like even when you're eating ramen noodles, you still gotta buy the ramen noodles.

Where do you get your ramen noodles from? Anyways, let's get away from the actual technology and let's talk about the FTC and Anthropic and OpenAI, because it looks like, looks like, maybe, perchance, probably not, but maybe, the government is actually going to come after Anthropic and OpenAI for the felonies that they have published on the New York Times. I always think about that. I took an Electronix crime class back in like 1998

So there was no Facebook, there was no Twitter. Friendster was still in the future. MySpace was still in the future. I think LiveJournal, LiveJournal might have been around at that time. But anyways, I still remember my Electronix crime professor talking about things that you, when you send digital communications of any kind, basically we're talking about email or ICQ or AIM back then, and basically he said that he goes by the New York Times rule. He does not, he did not say anything on electronic communication that he would not feel relatively comfortable with being published to the front page of the New York Times. Because as he said back then, wait for it, in 1998, anything that you transmit digitally basically can get out into the world. And so that is the funny thing we have now with Anthropic and OpenAI. I mean, they're taking that advice in just a very weird way.