What Do A.I. Companies Really Do With Data They Purchase?

Did you know that there is a company that auctions off the Slack histories and email histories of defunct companies? – What happens when the buyer actually gets a hold of the data? Like, where does it go? – So, basically, you use this data to rebuild whatever company — you’ve acquired their data, so you’re basically rebuilding an airline or an insurance company or a start-up as an environment for A.I. agents, as a training gym, essentially, for these agents to go out and try different tasks and see whether they succeed or fail. – You’re creating a nightmare parallel universe, where Spirit Airlines still exists and is booking flights. – Yes. Yeah, so you can kind of rebuild the company as a video game. You can sort of mine tasks from these emails and teams messages, and then you can actually see how things played out. So and if you. Can get Spirit Airlines to turn a profit in the sandbox, that’s AGI. Yes. So it’s basically you have these kind of data points that come from these companies about how people interact with systems, about how systems interact with each other, and about how customers, navigate through these sort of giant systems. Well. So what’s happening with these data sets, including the data set of the dearly departed Spirit Airlines, presumably, is that they are being turned into reinforcement learning environments for AI agents to learn new tasks. These models, now, the way that they are improved is through reinforcement learning. Reinforcement learning is sort of this trial and error process, where you go out, you sort of do a little task or a test or play a game, and you get a score or you get some indication of whether you’ve succeeded or not. And like, maybe you’ve broken into hugging face like says, yes, that one was a success. And then you sort of try it over and over again using slightly different strategies or slightly different techniques every time. And you sort of get signals about what works and what doesn’t. And that’s how you improve at things like autonomous coding. So for example, if you have the transaction history of an airline, you can say what happened. You know, in 2015 when there was a big storm on the East coast, how did the routing decisions get made? And did that result in people getting to their destinations on time? So there is a company, simple closure, whose whole business used to be sort of helping field startups wind down. But now. These guys are like the undertakers of Silicon. Valley. Yes, exactly. That corporate logo is just like the Grim Reaper. Yeah. Yeah. Yes. If these guys show up at your office or you get a call from them, it’s a very bad day for your company. So basically, they were sort of like, you know, helping do the orderly wind down of these things. But then they realized, oh, there’s actually a market for the data from these dead startups. And so they start selling it to AI companies.