The Robot That Might Fold Your Laundry or Kill Your Grandma

Meta just spent somewhere between $125 and $145 billion on data centers and AI research, and tucked inside that number, like a receipt you find in an old coat, is the acquisition of Assured Robot Intelligence, a company whose mission statement reads like a marketing intern discovered science fiction at 2 a.m. and decided to make it everyone's problem. Wang and Pinto, the founders, are real academics with real credentials, which makes it both more and less reassuring that they've joined something called Meta's Superintelligence Labs, a name that would be embarrassing if anyone at that pay grade were still capable of embarrassment.

What they're actually building, if you strip away the press release language, is a brain for a robot that doesn't exist yet. "Model capabilities for robot control, self-learning, and whole-body humanoid control" is the kind of phrase that sounds like progress until you sit with it long enough to realize they're describing the project of teaching a machine to exist in the physical world, which turns out to be genuinely, stubbornly, expensively hard. Nailing jelly to a wall is an understatement. The physical world is full of variables that don't care about your training data: a wet floor, a dog that moves unexpectedly, a laundry basket that's slightly more full than yesterday. Humans handle this without thinking because we've had millions of years of evolution quietly doing the math. A robot has a deadline and a VC-backed runway.

And I keep coming back to the old joke, the one where a guy walks into a bar and the bartender says sorry, we don't serve humans here, because it's both funny and not funny at all, which is the exact register in which most of the tech world operates right now. There's something genuinely useful buried in the idea of a robot that can help people who can't easily help themselves, an elderly person living alone, someone with limited mobility, the enormous invisible population of people for whom the gap between needing something and being able to get it is a daily, grinding fact of life. That's a real problem. That's worth working on. The question is whether Meta is working on that problem or working on the version of that problem that photographs well for an earnings call.

Because Meta's track record here is not exactly a confidence-builder (and I say this as someone who has watched them announce the future approximately fifteen times in the last decade, usually with a keynote that involves someone wearing a headset and gesturing at nothing). The metaverse was going to change everything. It changed the stock price, briefly, and then it mostly changed the definition of what counts as a sunk cost. Now it's humanoid robots and superintelligence labs, and the money keeps moving, and the press releases keep landing, and somewhere in a lab that costs more than most countries' GDP, two smart people are trying to figure out how to get a robot to fold a shirt without dislocating its own shoulder.

I don't think they're evil. I don't even think they're stupid. I think they're operating inside a system that rewards the announcement of ambition more than the delivery of it, a system where the distance between "we are working on" and "we have built" can be measured in years and billions and nobody really checks. Wang and Pinto probably believe in what they're doing. Most people who do this work do. The belief is not the problem. The problem is that the belief gets laundered through a company whose primary interest is in being perceived as the kind of company that believes in things, and those are not the same thing, not even close.

So we're along for the ride again, watching the numbers get announced, watching the lab names get more grandiose, watching the gap between the promise and the thing you can actually hold in your hands stay exactly as wide as it always was. Maybe they build the robot. Maybe it folds the laundry. Maybe it does something genuinely useful for someone who genuinely needs it, and that would be worth something, that would actually matter. Or maybe it ends up in the same drawer as the metaverse and the pivot to video and every other announcement that was going to change everything and mostly just changed the quarterly report. Either way, someone's going to get a press release about it. That part I'm certain of.