Chelsea Finn

The dream of deploying humanoid robots in every home ...

Community Perspective - Chelsea Finn - AI2050

Chelsea Finn, Stanford University

The dream of deploying humanoid robots in every home ...

Professor Chelsea Finn's Deep RL Course Now Online

Chelsea Finn on teaching robots to learn at AI Startup School

Chelsea Finn to speak on developing generalist robots at ...

Chelsea Finn

Interview with Andrew Ng and Chelsea Finn: AI & Robotics

Physical Intelligence (π)

Chelsea Finn: AI and Robotics Pioneer

Interview with Andrew Ng and Chelsea Finn: AI & Robotics

@blingdivinity @yoonholeee @roshen_nair @qizhengz_alex @Kangwook_Lee @lateinteraction https://t.co/4k1c7aTolt

Agents like Claude Code do amazing things relying critically on hand-engineered harnesses We show how agents can optimize the harness w.r.t. end performance Key idea: store all experience in a files

Usually, we expect more diverse data >> less diverse data. Cross-embodiment transfer seems to benefit from paired data across embodiments, more so than increasing diversity. Webpage & code

We lack simulated test-beds for studying memory in robot learning. RoboMME is provides a sim benchmark for multiple forms of memory. No single method performs well on all tasks! Paper, code, leaderb

Fun fact: Multiple of the tasks from Benji Holson's robot olympics challenge (https://t.co/ZwjSrkaLnt) worked poorly until we added memory. The policies for these tasks used an early version of these

We added short-term visual memory + long-term text memory to pi models. 🤖 Enables robots to: - complete tasks up to 15 min long - cook grilled cheese while keeping track of time - adapt in-context

Pi models are now running in production settings, in collab with @Ultraroboticsco and @weaverobotics. We see: - much higher autonomy with pi-0.6 over using pi-0.5 - fewer mistakes & higher throug

Video gen models make pretty videos, but lack physical accuracy Large robot data is helpful but insufficient, esp since this data is mostly demos By fine-tuning on policy data, we get far more accur