AI Labs Buying Customer Data from Dead Startups - Violating Privacy for Pennies

Eli the Computer Guy (0:00)

Oh, lordy, lordy, lordy. This could end up being pretty screwed up. But at least, at least, the one thing I have to say for Elon Musk is he's not actually trying to wipe out humanity.

I've been a bit of a Musk hater for a while now. I gotta admit it, right? I feel like he's been going in the wrong, wrong direction. But the one thing that I do have to say for Elon Musk is he may feel very comfortable killing poor black people, but he seems to have contained his willingness to kill people to poor black folks. Unlike Dario Amodi, who apparently is quite okay with just wiping out humanity.

Anyways, I just got done with a video talking about the Biolab that Anthropic set up, where they literally have come up with a variation of CRISPR. CRISPR, if you don't know, is a gene editing technology that allows you to do something called gene driving, where you make a modification of the DNA of an organism, and that then gets reproduced into the organism's offspring.

Yeah, yeah, that's what Anthropic is up to right now. So, look, let's be honest.

Elon Musk, at this point in time, almost gets a completely and utterly free pass. Like, whatever the hell Elon Musk is doing. Oh, Elon Musk is gonna poison poor black people? I mean, that's horrible. But, you know, he's not trying to gene drive into human DNA, so, you know, there's that. Oh, he wants to launch a rocket ship, like, every hour? Which will just completely destroy the environment of Louisiana? Well, at least he's only destroying the environment of Louisiana. He's not trying to, like, destroy the world.

Hey, I don't know if Elon Musk is the hero of this story, but at least he's no longer the worst villain, I suppose.

I hate this timeline. Anyways, Elon Musk, coming from Gizmodo, Elon Musk reportedly seeking customer data from dead startups to train AI. All the data from X wasn't enough to quench Grok's thirst for AGI. So I do kind of think this is interesting as we move forward with artificial intelligence. There is this idea with the current paradigm of artificial intelligence, and you do have to understand it is a current paradigm. There have been multiple other paradigms for artificial intelligence. So this whole transformer LLM model thing is just simply the current paradigm. You have Jan LeCun over, I think he's in France right now. He's actually coming up with his own entirely new paradigm. So it is important to understand. We talk about artificial intelligence. Again, that's not really a word. That's just like the psychosis of Gen Xers remembering 1980s sci-fi movies and thinking they were far better than they actually were. Right, AI is a word. But the thing is with this current paradigm of artificial intelligence, the idea is the more data, the more intelligence, right? And that's why you have OpenAI and all these companies, and they're just scraping all the data. They scrape the entire internet. You have companies now basically buying rare books, cutting the spines off of them in order to ingest those in their AIs. You have these companies, they're going out there, they're scraping reddit and they're scraping twitter, because literally, their argument is that their AI has to understand real human communication, and what is more human communication-y than reddit? This is how fucked up these people are. Again, I talk about how demented and senile and psychotic the CEOs and founders of these frontier labs are.

I think that actually shows you, right? That is the proof of how just out of touch these creatures are, right? They think reddit is normal human communication. Who else in the entire world, even reddit users, even reddit users, don't fucking think reddit is normal human communication. But these people are just so bat crap crazy, right? Anyways, scraping reddit, scraping X, there's always this thing, more data, more data, right? We saw this with meta. Meta was going to start tracking everything that all of their employees were doing on the computer, more data. And so we're currently in this paradigm. And so basically the thought here is, is like now that they scraped the entire internet, now that they scraped reddit, now that they scraped X, now that they've chopped, destroyed books in order to ingest the books, where are they going to get more data from? And so the idea is to get data from dead startups. And so this could be relatively concerning, because one of the big problems with a lot of these startup companies is they don't necessarily know what they're doing to begin with, right? Security procedures, anonymization, hashing, there's a whole bunch of stuff that goes into it. Collecting way too much data, all of that kind of stuff. And so one of the interesting things is if the startup dies, so like most of the times when startup dies, right, they may auction off their servers or equipment or whatever else, but a lot of the data, there's a lot of question about the value of the data. And even if there is a value in the data, normally it's a subset value of the data. Like what products were customers buying? How did customers or users move through the website? Kind of like relatively specific things. Like, okay, like we, so we are an old school company, there's a startup company that is trying to compete with us in our space, they fail, but they learn a lot of lessons, right? That's a big thing about startup companies is they spend a lot of money to learn how not to do things. So if you can buy their data to learn on the cheap, what also not to do, that can be a great thing. But again, the data that you're going to be looking for is going to be subsets of the information. It's literally not going to be essentially the entire database itself. You're going to be looking for very specific things. The interesting thing now is you have these AI companies, again, they're desperate for data. So now what they want to do is they want to ingest all of the startup's data. And when we talk about this, like you think about that, like, yeah, yeah, what does that mean? A lot, holy fuck, it could be a lot. And again, the whole question, we begin with things like AI security. When we see how bad these companies are, we create AI agents and the rest of it. So when they say they're going to anonymize, anonymize information, if they're so pissed poor, they don't have real time visibility into what their AI agents are actually doing at any point in time. Do you really think they have the required skill sets to fully anonymize data that was never meant to be anonymized? That's one thing to understand. Sometimes you get data sets that have identifiable information, just kind of sort of because. And so you go through and you anonymize the data. But it's a data set that was supposed to be released, it was supposed to be exported, that just happens to have too much information. But think about this, what happens when you're ingesting just pure email, pure emails of normal conversations? What happens when you're ingesting teams' conversations? What happens when you're ingesting Slack conversations? What happens when you're ingesting just pure documents? What happens when you're ingesting whatever the hell is in their database system? And you're just ingesting it, right? Do we really trust these companies to be able to anonymize the data?