arstechnica ,
@arstechnica@mastodon.social avatar

What kind of bug would make machine learning suddenly 40% worse at NetHack?

One day, a roguelike-playing system just kept biffing it, for celestial reasons.

https://arstechnica.com/gaming/2024/06/what-kind-of-bug-would-make-machine-learning-suddenly-40-worse-at-nethack/?utm_brand=arstechnica&utm_social-type=owned&utm_source=mastodon&utm_medium=social

langeweile ,
@langeweile@social.bau-ha.us avatar
nazokiyoubinbou ,
@nazokiyoubinbou@mastodon.social avatar

@arstechnica This is the fundamental problem with "AI." It's not AI. It can't learn. They have to manually program those conditions in.

The fundamental problem is that in the real world "full moons" happen frequently. Both literally and figuratively of course.

axleyjc ,
@axleyjc@federate.social avatar

@arstechnica
Ooh, they added an LLM?

Klassika ,
@Klassika@mastodon.social avatar

@arstechnica
I couldn’t even play NetHack on my first PC. It had a 8088 as a processor that couldn’t manage the lunar cycle, or so I wad told.

rbjolly ,
@rbjolly@vivaldi.net avatar

@arstechnica You say "bug" like it is a bad thing. ; )

  • All
  • Subscribed
  • Moderated
  • Favorites
  • random
  • All magazines