Geoffrey Hinton, the recent Nobel Prize winner who worked at Google's Deep Mind for years, warns us about a specific kind of dumbing down from AI, and he tells us exactly how it happens.
AI is trained on canned datasets. They come from the data lakes of the major cloud providers (Google, Microsoft, Amazon). AI doesn't scrape the internet directly, it buys internet datasets from the people who collect the data (like Google). So whatever is in those datasets, right or wrong, is what goes into the AI.
Then the kids use AI like an encyclopedia, and everything they learn goes right back out to the internet, on Facebook, TikTok, and chat groups. Which is then picked up the next time Google scrapes the internet, and then it goes right back into the AI.
The result is that old outdated information ends up lasting a lot longer than it should. The ratio of information consumption to information production goes up, and there isn't even an editorial board like Wikipedia, all there is are political censors.
Unfortunately for all of us, Google doesn't make any money from ethical considerations. They make money by scraping the internet, and providing the results to AI. They want the dumbing down cycle to continue, because they make money off it. The dumber people get, the more they use AI, and therefore the more money Google makes.
shumer.dev
