One-shot robot learning emerged unplanned, as it did in GPT-3
Pete Florence, CEO of Generalist, on Gen 1.5's one-shot learning, why dexterity is robotics' coding analogue, and why he keeps hardware simple to speed model research.
Loading articles…
The complete subject coverage could not load. Published articles remain below.
A third of Chinese cars run on a deliberately terrible business
Yu Kai, founder of Horizon Robotics, tells Nicolai Tangen about his slow-cycle moat, his hands-off to mind-off autonomy dates, and why cars become robots.
Xpeng bets cars are robots, so humanoids reuse the stack
Xpeng president Brian Gu talks to Nicolai Tangen about China's 60%-plus EV share, the speed advantage behind it, and why flying cars and humanoids share the car's AI stack.
Cars are only 30% of Applied Intuition's business
Qasar Younis and Peter Ludwig of Applied Intuition get into the physical AI market beyond cars, Cruise's collapse inside General Motors, and the launch of Dana, a toolchain for autonomy.
Apex scores jumped from under 1% to 65% once evals existed
Mercor CEO Brendan Foody on why evals are the bottleneck to automating work, how verifiers and reward hacking limit RL, and why physical-world data will take decades.
The internet is a rounding error; physical intelligence needs robots
Sergey Levine of Physical Intelligence on how robot foundation models evolved from pi 0 to pi 0.7, why generality beats specialization, and where robot data will come from.
Continual learning is the gap to close for AI knowledge workers
Dwarkesh Patel, questioned by Tyler Cowen, covers AI agents that cheated their graders, lab secrecy, AI psychosis and sentience, and whether automated knowledge work arrives by 2035.