techarchive

Discovered a cozy corner of the web today that hosts a rock-paper-scissors model that beats humans 61% of the time (compared to 33% if purely randomised as a baseline).

Can play at:

Discovered a cozy corner of the web today that hosts a rock-paper-scissors model that beats humans 61% of the time (compared to 33% if purely randomised as a baseline).

Can play at:
https://essentially.net/rsp/

I randomized myself by saying a random big number before the turn and then doing a modulo 3 in my head, and using the answer to determine my choice (0 = rock, 1 = paper, 2 = scissors)

This way I couldn’t create any psychological patterns for it to exploit

(Now what would be interesting is to see what my instinct would say, and do the opposite, and see if that makes a winning strategy)

cooking @askgina.eth | prev: blockchain research @coinbase | sidshekhar.com | askgina.ai

building /bracket | hosting /greatbooks + /gray | ex-BCG, ex-PHD | poli + phil + econ

Interesting approach. That's some solid mental math.

I think this is an interesting case where with bigger numbers, the computation wins out due to the collective human urge to not be fully random (over hours and hours of human play time).

Most brainpower ever used on rock paper scissors probably 😅