News for hackers. Stuff that matters.

Ask HN: What's stopping us from putting agents in general purpose robots?

After using Claude a bunch in the last few months I can't stop thinking the obvious next logical step is to put an agent in a physical robot

For example imagine asking "can you go into the other room, find object X, and bring it to me"

It seems to me the intelligence/reasoning ability is more than enough to do simple tasks

And giving models images they seem to be able to recognize objects well to understand what they are looking at

But its not clear to me what is holding this product area back.

Is it a matter of the speed of the update loop being too slow to be practical?

Or does interpreting images/video cause the context/cost to blow up?

( no comments | on HN )

There are no comments yet.