Three Times A.I. Was Just People
A.I. can seem like a Frankenstein’s monster of modern technology, a looming threat poised to take our jobs and our women and then kill us all, but the reality is way sillier. Not only is A.I. prone to all kinds of hilarious errors, they’re often being made by just some dude.
Amazon Fresh’s ‘Just Walk Out’ Technology Was Just Hundreds of Indian Workers
In an attempt to rationalize regressing to brick-and-mortar stores, Amazon claimed its Amazon Fresh locations were equipped with “Just Walk Out” technology that would allow customers to, you know, do that while an unseen entity totals up their purchases and charges them automatically. The only problem was that’s not actually possible, which customers started to suspect when they weren’t getting receipts until hours later.
This article not your thing? Try these...
What was actually happening was a team of 1,000 Indian workers watching videos and manually tallying customers’ orders. Amazon has since shuttered the “Just Walk Out” program in favor of scanners inside shopping carts that kind of work like a mobile self-checkout, because the problem with self-checkout lanes is that they don’t move enough.
Presto Automation’s Fast-Food A.I. Was Mostly Filipino Workers
In what must have seemed like the perfect evil business plan, Presto Automation offered fast-food restaurants like Carl’s Jr. and Del Taco technology that could take customers’ orders without the need for demanding human workers. The description sounds a lot like “a touchscreen,” but Presto Automation insisted its service used artificial intelligence. They changed their tune, however, when the Securities and Exchange Commission started sniffing around and the company was forced to admit that more than 70 percent of orders were being taken by human workers in the Philippines. In other words, regular old real intelligence.
ChatGPT Was Cleaned Up By Low-Paid Kenyan Workers
ChatGPT seemed to be a technological revolution, an A.I. that could absorb an entire internet worth of written word and parrot it back to us. Unfortunately, the internet is terrible, so initially, ChatGPT was specifically the parrot owned by your racist uncle. To stop it from trying to indoctrinate users into anti-Semitic cults and tell them illegally dirty jokes, OpenAI recruited an army of Kenyan workers, some paid as little as $2 an hour, to retrain ChatGPT to filter out the “horror show” parts of the human canon that it couldn’t parse without them.
It turns out the enemy trying to take our jobs was just old-fashioned international outsourcing, except it’s truer now than ever that they aren’t jobs anyone would want.