Wrote some thoughts on the scale of real data we need for a GPT in robotics, the cost of collecting that data, how surrogate data can reduce that cost, the drawbacks of surrogate data, and how Iโd think about reaching 1.1 trillion tokens.
https://t.co/Q6eVUmrEBB
We're nearing the GPT moment for robotics.
But to get there, we need models that combine broad generalization with high task success.
Thatโs why Iโm covering @physical_int โย the company building a foundation model for robotic intelligence โ in this edition of Startups to Join.