Your sexy AI girlfriend is really an underpaid guy in Kenya

One of the first steps in an effort to organize workers in Africa and South America is finding out exactly what those workers do for massive international tech firms. The answer turns out to be: Pretend to be sexy robot ladies and teach supposedly all-powerful AIs how to read.

Content moderation for Facebook? A lot of that work falls on women in Nairobi who frequently come face-to-face with violent, racist, sexist posts that leave lasting emotional scars. Marking up human-generated text so it can be added to AI training data? That's workers in Lebanon facing rules that prevent them from looking for other jobs, negotiating for higher wages, or even having a bank account. Adding those same markups to user queries? That seems to be happening in any place where data workers can be hired cheaply.

And when it comes to those always understanding, ever there for you, and reliably horny AI girlfriends ... those are just guys in Kenya texting directly with users while pretending to be sex robots.

Chat moderators are hired by companies such as Texting Factory, Cloudworkers, and New Media Services to impersonate fabricated identities, often romantic or sexual, and chat with paying users who believe they’re forming genuine connections. The goal is to keep users engaged, meet message quotas, and never reveal who you really are. It’s work that demands constant emotional performance: pretending to be someone you’re not, feeling what you don’t feel, and expressing affection you don’t mean.

How many other AI wonders are really low-wage workers being hidden behind a big tech curtain? That's hard to say. But stories of workers being called on to perform jobs that many assumed were done by automated systems are pervasive. So are tales of data workers being glued to their keyboards by the same kind of trap that keeps Uber drivers fighting for just one more ride.

Another highly visible AI product took a literal public fall this week with a viral video showing one of Tesla's "Optimus" robots hitting the ground during a public demonstration.

The gesture the robot performs right before it topples backwards shows that it is mimicking the actions of a human operator, right down to following that operator's hands as they pull off a control headset. Once the operator checks out, the only thing in control is gravity.

When Elon Musk first introduced the Tesla robot in 2021, the moment was incredibly embarrassing. The announcement was nothing more than a man dancing in a robot suit. It's one of the most cringe-worthy things that ever happened at a Tesla event, and that's saying a lot.

A year ago, Musk's big "Robotaxi" event also provided a chance to show off his Tesla bots. As Fortune reported:

Elon Musk’s humanoid robot wowed attendees as a half-dozen or so navigated their way safely through the crowd at Warner Bros. studio lot in Hollywood entirely untethered.

Just two years ago, the droid couldn’t even make it onto a stage on its own. But the latest iteration was talking with guests, playing rock-paper-scissors. or simply pouring a drink from a tap for the thirsty—before coolly flashing a peace sign with its metallic hand.

“This will be the biggest product ever—of any kind,” Musk predicted.

But the reality was a lot less impressive than Musk made it seem.

They were voiced and operated remotely by humans wearing special suits that translated their movements to the droid.

That these robots were being operated by humans was clear to any astute observer, but Daily Grail reported at the time:

... many of these ‘tech guys’ who were invited to the event seemed to fall hook, line and sinker for it – no doubt a consequence of having drunk a little too much of the Tesla Kool-Aid, given most were investors or fan-bois.

Those drinking that Kool-Aid include financial analysts who have valued just the Optimus portion of Tesla's business as high as $800 billion. Musk has said that Optimus will "be 80% of Tesla's business."

Almost five years after it was introduced, Optimus has graduated from man-in-suit to a very large radio-controlled toy. What does this say about Musk's RoboTaxi? How many of its miles are really controlled by operators hidden away to make it seem more capable?

There's no doubt that AI remains a very serious threat. The massive investments in this technology are already distorting the stock market and eliminating tens of thousands of jobs. Both the people who are making the AI models and the researchers who cover the field believe there's a very real threat that AI will be hugely disruptive, and even actively destructive, in ways we're not prepared to address.

But there's also no doubt that for every genuine AI advance, there's an equal or greater amount of pure marketing hooey. Companies aren't just racing each other for the elusive "artificial general intelligence" (AGI) crown; they're competing for investment dollars they can burn in the furnace of spectacularly costly creations that have no clear path to ever making a profit.

AI consumes huge quantities of electricity and water, but it's really fueled by hype. Lots, and lots, and lots of hype produced by men whose jobs have always been making meager achievements sound revolutionary. The best thing that could happen for the economy and for AI research is for investors to wake up, smell the nonexistent coffee, and start putting on the brakes.

Otherwise, the whole world is going to fall on its face, or back, while pretending human operators are no longer needed.