General Intuition Brings Gameplay-Trained Models to Robot Control
Sign in✓ Signed in
General Intuition is using action-labeled video-game footage to train models that navigate virtual environments and control a quadruped robot, adding a physical-world angle to the foundation-model race. A public demonstration required eight minutes of real-world robot data after pretraining, but it remains a narrow transfer result rather than evidence of general-purpose autonomy.
Continue reading
The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in✓ Signed in — this article isn’t included in your current plan.
Unlocking the full article...