Actually flipping it is still well beyond the state of the art in robotics... but maybe it’s important to avoid “the superhuman human fallacy” of attributing competence to e.g. me that humans don’t possess
Conversation
Replying to
It’s interesting. A specific pancake-flipping robot would likely be possible. A generalized robot capable of dealing with a large class of such tasks would likely be intractable. Humans seems to learn a semi-generalization around prototypes based on limited induction.
Did you ever read the qualitative physics debate papers? One of my favorites
2
2
I’m pretty good at unsticking and flipping things that need it. The trick is to let the oil, temperature level, and covering/uncovering with a lid do most of the work. The mechanical part is the hardest but can be rendered trivial.
1
2
Show replies
Replying to
Yes, you beat me to it!
Quote Tweet
Replying to @ndwpdx
Yes… at the time I was interested in general-purpose human-like vision. For that, one-shot learning is a consideration (among many others)
meaningness.com/eggplant/ratio
2

