This is a single uncut video, showing a robot learning several tasks instantly, after just one demonstration each ...

This is possible because we've now been able to achieve in-context learning for everyday robotics tasks, and I'm very excited to announce our latest paper:

🎆 Instant Policy: In-Context Imitation Learning via Graph Diffusion 🎆

robot-learning.uk/instant-policy

(1/6) 🧵👇 In-context learning is where a trained model accepts examples of a new task (the "context") at its input, and can then make predictions for that same task given a novel instance of it, without any further training or weight updates.

Achieving this in robotics is very exciting: with Instant Policy, we can now provide one or a few demonstrations (the "context"), and the robot instantly learns a closed-loop policy for that task, which it can then immediately perform.

(2/6)

Want to teach your robot new tasks from only a single demo?

We've just released code for MILES, which we presented at CoRL 2024 last week.

Learning is fully automated: you just provide a single demonstration, then sit back and relax! 🍹😴

Code: .

🧵👇 robot-learning.uk/miles Once you have the code, it's very easy to teach your robot.

You provide one demo + one reset, and then the robot collects its own self-supervised data.

This is much easier than reinforcement learning, because you don't have to keep resetting the environment after each episode.