AI “learns” to better understand and ultimately mimic us by scraping our behavioral and content data from the Internet.
It’s a euphemism for stealing.
I’m not talking only about copyrighted material, though that’s certainly a part of it. Sarah Silverman is suing Meta and OpenAI for consuming her memoir. You have to imagine there are zillions of other instances of AI violating copyright to “train” its LLM models (Getty is trying to protect its images with a similar lawsuit).
There’s another euphemism for you.
I’m also not focused solely on all of the data that we give away to companies when we sign their opaque mouseprint “terms of service” agreements. Google’s AI gets to read every email it transmits for us and Facebook and Twitter’s Ads do the same with our posts. Amazon uses our shopping history to teach its AI how to better think like us, though I don’t remember ever giving it that authority.
Surveillance is often implicitly allowed because it’s not overtly prohibited.
Nor am I interested only in the human employees who are “working side-by-side” with robots to help train them how to do their jobs, from factory assembly lines to steering cars. Once the skills transfers are complete, the AI will be able to assume those jobs.
No, I’m talking about the whole shebang, the Big Kahuna of 3.5 quintillion bytes of data that are created every day working, browsing, chatting, uploading and downloading, and playing games. We have been encouraged to migrate our lives online in exchange for convenience.
For every miracle of medicine or professional opportunity provided by digital tools there are many thousands of instances in which people just found a pair of shoes they wanted to buy, or found themselves compelled to insult an utter stranger.
By making our lives more convenient, the Internet has made it easier for AI to study us.
Think about that for a moment. Generative AI is able to mimic how we think and communicate because we make it possible for it to analyze how we think and communicate. Every day, we produce more data so that it can become more like us.
The charitable and perhaps mainstream view is that it’ll give us even greater conveniences along with the occasional meaningful insight breakthrough.
A more cynical view would be that it’s kinda like a slow walk Invasion of the Body Snatchers. AI models are harvesting our humanity to become us, at least when it comes to doing our jobs.
There’s no copyright or regulation that can help us. We are giving up what it means to be human.