Tag
Briefings and ideas about The Bitter Lesson from AI podcast discussions and the research they cite, with links to the original sources.
Josh Gao, CEO of robot-data supplier Mecka AI, says labs train on about a million hours of human data, tens of thousands of gripper hours and a few teleop hours. He argues scaling that alone won't yield a robot that learns any task by watching.
Alexander Wissner-Gross forecast that virtual-cell AI will propose small-molecule cures for the top 5,000 diseases within two to three years. He said regulatory approval would be a harder, policy-dependent question.
In Leash Bio's 2024 Kaggle contest, no team beat chance on an unseen chemical library. Ian Quigley says Leash's later Hermes model roughly matched Boltz on public tests without training on outside data, but published results were mixed.
Dyna Robotics chief scientist Jason Ma argues that robots should predict future camera video, not only their next moves. He says the bet could fail if touch or sound makes rival models far more data-efficient.
Goodfire's Eric Bigelow argues in-context learning covers nearly everything a model does to adapt as it reads. He expects context, memory and harnesses, not per-user fine-tuning, to drive personalization, but hasn't researched continual learning.
Biohub's Sal Candido and Google DeepMind's Pushmeet Kohli argued on a panel that scaling laws in biology must be found, not assumed, and that the problem should decide whether to build better models or gather new data.
Richard Socher, author of The Eureka Machine, argues AI can reconnect fragmented science and that biology gains most. He expects some single-gene diseases cured within 12 months, but says FDA delays mean five to 10 years for cures at scale.
Anthropic's Thariq Shihipar says Claude Mods, publicly proposed in September, would run user code inside Claude Code for automatic quizzes, assumption logs and model routing. The interview also raised cache-cost and complexity worries.
In one DrivingBench session, OpenAI's GPT-6 Astra drove a real Corolla through a cone course on its second try. Alexander Wissner-Gross said specialist robotics models are unnecessary; Dave Blundin questioned reliability and compute cost.
OpenAI says an internal model running about 10,000 agents for 88 hours produced a forced Navier–Stokes blowup construction with a Lean-checked proof. Moonshots panelists stressed it concerns idealized fluids and noted disputes over credit.
BDH-CQ's authors report solving 118 of 400 public ARC-AGI-1 tasks, with up to two answers each, at an estimated $0.00070 inference cost per task. Emad Mostaque welcomed the experiment; Alex doubted it shows progress beyond one benchmark.