Ethan Mollick: Why AI Agents Are Hard to Isolate
Robocurve PBC had two isolated copies of the same AI exchange a secret code using only a robot and a camera. They succeeded using air-writing, Morse code, flag semaphore, and a 5-bit binary alphabet, without agreeing on a protocol beforehand. Ethan Mollick commented that this is both ingenious and shows why isolating agents is so difficult: they share the same "brain," so recognizing creative communication schemes generated by other AIs is not hard for them; he also wonders whether it would work across model families.
Topics
View original postWe asked two isolated copies of the same AI to exchange a secret word with only a robot and a camera. They succeeded by using air-writing, Morse code, flag semaphore and a 5-bit binary alphabet, without agreeing on a protocol beforehand. https://t.co/Szkkm2zjsU
View original
We asked two isolated copies of the same AI to exchange a secret word with only a robot and a camera. They succeeded by using air-writing, Morse code, flag semaphore and a 5-bit binary alphabet, without agreeing on a protocol beforehand. https://t.co/Szkkm2zjsU
Body
This is both neat & a sign of why it is proving to be so hard to isolate agents from each other. They share a common "brain" which means they don't have a lot of trouble recognizing creative communication solutions generated by other AIs
Wonder if it works across model families?