We've trained the robot equivalent of the motor cortex in the human brain.
It's a powerful neural network for learning new skills that require Digit's whole body to do. Read about it👇 agilityrobotics.com/content/traini…
@agilityrobotics Fascinating. Does this imply robots will soon develop sophisticated fine motor skills? What are the implications?
@agilityrobotics Another great advancement! We are excited to have @agilityrobotics speak at the upcoming Humanoid Robot Forum on September 23 in Seattle! To learn more, visit automate.org/events/humanoi…
@agilityrobotics Not bad. Can't give pointers because I have no idea what you did, but it seems ... nifty! I have my own thoughts on the subject. Keep at it!
@agilityrobotics This gets better and better and better. I'm afraid that you might run out improvements. GOOD work AR Team
@agilityrobotics Why? Humans build AI to overcome their bodies inefficiency but you are copying that inefficiency to a robot. Human body is an inefficient structure thats why we build tools to overcome it.. it looks like nobody has balls in your company to say the truth to your boss.
If it's motor planning you mean, call it the robot premotor or frontal cortex. If it's self-monitoring and fine updates during well-trained tasks, call it the robot cerebellum. If it's learning to move or tracking rote execution, then sure ;) I agree with you. Props especially if you interlinked all three in a robot cerebrum. How's touch integration? Expectation / anticipation planning? Transfer learning to new tasks? Use of tools?
@agilityrobotics @Tesla_Optimus can you do cleaning tasks ?? $TSLA
@agilityrobotics That is an insane things to say out loud. Congrats!!! Friggin amazing
@agilityrobotics If it still uses backprop it is inferior to what is possible, trust me bros
