Microsoft New Surface Laptop Ultra with NVIDIA GPU for Local AI - OpenAI is DEAD

Eli the Computer Guy (0:00)

Okay, boys and girls, let's say it all together, OpenAI is dead. That is right. This is the very long form explanation as to why OpenAI is entirely screwed. I am going to be doing a 5,000 part YouTube series on why OpenAI is the technical term, very, very, very, very dorked. So I think this is interesting coming from Microsoft, or actually coming from CNBC, Microsoft to sell $2,599 Surface Laptop Ultra containing Nvidia AI chip. So basically, this laptop is going to contain the same chip that the DJI Spark, anyway, the Spark computer from Nvidia has. And so the idea here is that Microsoft is trying to build laptops that can actually do relatively powerful local artificial intelligence processing. And this is the path that we have seen for a long time. So one of the things that I've been talking about for a while is I honestly believe that the thesis, the fundamental thesis for OpenAI is utterly incorrect. If you go and you read books about OpenAI and you listen to Sam Altman and you see what OpenAI is doing, what you will see is they have a very strong thesis that they are basically bound to at this point in time. And basically what that thesis is, is that more data and more GPU, more hardware resources gets you more AI and that AI should be centralized like a utility like water or electricity or those types of things. And so the concept there is that if OpenAI can gobble up all the resources and it can gobble up all of the data, then it can create the best AI system and then everybody will be forced to use their AI system. And basically it'll be a metered monopoly of some sort. And one of the things that I've argued for a long time is, I just think that's wrong.

I just think it's a fundamentally wrong thesis. One of the things I've been screaming about for a while now, the things that I like, is small language models. I think small language models are amazing. Small language models are incredibly small models that you can run on almost anything. And one of the things that I argue is that for artificial intelligence, or really technology in general, the point of technology is to solve problems. As I've talked about before, you do not need a Maserati to go to the grocery store. You don't need a Maserati to commute to work in the morning. If people told you, Eli, Eli, don't you understand how much better a Maserati is than your Ford Escape? And the problem is, I mean, technically I guess they're probably correct, right? The Maserati has better engineering, the Maserati has better materials, the Maserati has a lot of things that are much better than my Ford Escape. But my Ford Escape cost me about $40,000 about five years ago, and yeah, I just guess that's what I needed to do. I don't need a Maserati. I don't care how much better that Maserati is. It's actually not going to get me to the grocery store any faster. Here's one of the things I think a lot of people don't understand. It does not matter how expensive your supercar is. If the speed limit is 25 miles an hour, we're all going 25 miles an hour, right? Literally, the guy on the e-bike might beat your million dollar fucking supercar in the real world, right? And that's the thing, as technology professionals, we're actually looking for solutions. And that's where I talk about things like small language models, where if you actually look at what AI is good for, what the AI stack is good for, I talk about this with language processing and parsing. There's a lot of specific tasks that this technology is good for. And what you realize is a lot of the small language models actually do that very well. And so with these small language models, being able to do things very well, we've seen other companies come out and start creating products. So one of the things I've talked about before is like with Google. So Google now builds an LLM directly into the Android operating system. Google now has an LLM directly for the web browser, for Chrome browser. Google has an LLM that can run on any laptop that has over 16 gigs of RAM. And obviously, Google has a Gemini which you can connect to online, right? And so that's what we're starting to see is the architecture of how things always get built. This isn't the architecture of how AI gets built. This is the architecture of how every fucking thing in technology gets built over the past 30 years. Because we've actually figured out systems architecture over time. You know, whatever it is, whatever it is we are trying to accomplish. What we've found is there's reasons for doing things. There's reasons to centralize. There's reasons to decentralize. And again, whether it's a database system, whether it's an artificial intelligence system, whether it's a gaming system, whatever it is, we kind of know how you should build those things. So one of the things I've been arguing for a while now is having one centralized monolithic system, essentially for everybody, is just fundamentally not how it's done. Again, I don't care if it is Sam Altman or the Dalai Lama running this particular technology company, right? The fact of the matter, that is just simply, it's not about the morals or I mean with Sam Altman, it's a little bit about the morals and ethics. He is just a god-awful, horrible human being. But anyway, but really it's not about the morals and ethics of who's running the company. It's just the concept, the thesis fundamentally doesn't make sense. And I think that is where we're going to start to see a lot of problems for OpenAI and Anthropic and such coming down the pike, especially as their competitors start coming for them. One of the weird things that I've seen, I've been in technology since 1996, over 30 years now. And one of the things, good and bad, good and bad, sometimes you're eating your popcorn as you get punched in the face. You know, that's how our field is. Is that, you know, basically this is an industry of type A autistic assholes, generally. And so one of the things you have to understand is in our industry, is like people fuck with each other just to fuck with it. It's not even they fuck with each other just because they can. I mean, that is a thing. No, sometimes you see fucking interns trying to fuck with CEOs for some reason. You're like, what the fuck is wrong with you?