During a visit to Generalist AI in Cambridge, Massachusetts, I witnessed robots performing household tasks with surprising dexterity. These weren’t just pre-programmed movements; they were learning on the fly.

The star of the show was a robot that swept a block into a bowl using a dustpan when its brush was removed, demonstrating its ability to improvise and adapt. Another two-armed robot unzipped a purse and retrieved money, then switched hands for better grip—much like children figuring things out.

Generalist is focused on teaching robots the physics of our world, which may explain their impressive adaptability. The company has amassed a vast amount of high-quality training data and built its AI models from scratch, setting it apart from competitors.

The cofounders—Pete Florence, Andrew Barry, and Andy Zeng—have impressive backgrounds in robotics and AI research, previously working at Google DeepMind and Boston Dynamics. Their goal is to deploy these learning robots in real-world commercial settings, though their current success rate of 59% for completing tasks still has room for improvement.

The potential for rapid skill acquisition in manufacturing is immense, but only time will tell if Generalist’s approach can truly revolutionise the way we do things around the house and on the factory floor.