The North Pole’s Dirty Secrets and the Creepy Rise of Humanoid Data

I’ve been sitting on this edition of The Download from MIT Tech Review for a bit, and two stories jumped out at me as worth chewing on. Not because they’re breaking news, but because they quietly point at something bigger.

First, the North Pole.

Last year, a research vessel sailed straight to the North Pole through open water and thin ice. That’s not supposed to be easy. In the old days, you’d need an icebreaker and a lot of patience. Now it’s practically a leisure cruise. That’s a stark reminder that the Arctic is changing faster than most models predicted.

So scientists are doing something smart: they’re drilling deep into the seabed to look for clues about whether the Arctic Ocean was ever completely ice-free in the past. If it was, what did that look like? What happened to the ecosystems, the currents, the weather patterns? And more importantly, what does that tell us about where we’re headed?

This is the kind of long-term science I wish got more attention. Not flashy, not funded by a VC with a timeline, but essential. Tim Kalvelage wrote it up for the print edition, and it’s worth your time.

Then there’s the humanoid data thing.

James O’Donnell got invited to an app that pays people to film themselves doing mundane tasks: putting food in a bowl, microwaving it, opening a drawer. Another site wanted him to remotely control a robotic arm to help improve its dexterity.

This is the weird, messy underbelly of the humanoid robot boom. Everyone’s racing to train robots to move like us, but the data is hard to get. So companies are turning to gig workers and everyday people, turning our most boring movements into training data for the next generation of machines.

It’s a little creepy when you think about it. But also kind of brilliant. And it’s one of the 10 Things That Matter in AI Right Now, a new series from MIT that’s worth tracking.

The rest of the roundup is the usual chaos: Google, Microsoft, Amazon, and Meta all set AI spending records, collectively up 71% from last year. Microsoft and Google are seeing payoffs, but Meta’s stock slid because investors got spooked. The White House is opposing Anthropic’s plan to expand access to its Mythos model, citing cyber risks. And Elon Musk testified that OpenAI’s leaders “looted the nonprofit” and that he “was a fool” for trusting them.

Same circus, different day.

But the two stories that stuck with me are the Arctic drilling and the humanoid data. One is about understanding the past to prepare for the future. The other is about selling the present to build the future. Both are worth thinking about.