BLUE
Profile banner
MA
Michael A Osborne
@maosbot.bsky.social
Dad, husband, dog dad, Australian, Professor of Machine Learning in Oxford. Bayesian ML, Long Covid, porridge, AI must be good for humans, www.robots.ox.ac.uk/~mosb
3.2k followers304 following563 posts
MAmaosbot.bsky.social

DON'T: use bugs like "r's in strawberry" to argue that "AI is slop", as any relevant bugs are very likely to be patched as AI advances DO: use these bugs to point out that AI is inhuman, and that we don't know how many more such bugs lurk within its opaque architectures

This image shows a question asked in a ChatGPT interface, using model o1-preview. The user asks, “How many r’s in ‘strawberry’?” The AI responds with: “There are three ‘r’s in the word ‘strawberry’,”
2

Ppookleblinky.bsky.social

Not quite an advance as an if statement to hardcode that answer for that question. Duct tape on a dead horse, really.

1
MNastro-markus.bsky.social

How many lights do you see? - There are four lights. Jean-Luc Picard - Chain of command

0
Profile banner
MA
Michael A Osborne
@maosbot.bsky.social
Dad, husband, dog dad, Australian, Professor of Machine Learning in Oxford. Bayesian ML, Long Covid, porridge, AI must be good for humans, www.robots.ox.ac.uk/~mosb
3.2k followers304 following563 posts