Fable's safety filters block benign math and nonsensical queries, prompting jokes from researchers
The blocked prompts cited cybersecurity and biology policies.
Negative users called Fable AI's safety filters ridiculous and pretentious for blocking math queries and other benign topics like cybersecurity.
Most Activity
Good news everyone, kernel methods are AGI, pack it up
@jon_barron 😂😂😂
@liqsweep Fable 5 asked to ask them about mantis shrimps
Opus 4 asked
Fable 5 started writing about mantis shrimps, and the classifier started firing
How does this even work
@rbrthogan I choose to believe that it's actually excellent at safety filtering and that every 1980s math thing I'm interested in is critical to AGI.
@jon_barron :)
@jon_barron They knew you were Jon Barron somehow
@jon_barron Kernel methods or otherwise, it seems safe to assume that any company deciding what's relevant for future AI developments is pretentious BS.
@jon_barron The real reason they wanted to hide it from us is because of how ridiculous it is to call this safety.
@Lari_island @liqsweep Maybe because of biology? They can't talk about biology right now, the rails cover the entire category. I got a refusal on "epigenetics" after they introduced "genome" themselves. I've read that biology, chemistry, and cybersecurity are currently blanket forbidden
@jon_barron I guess it's your IP lol
@Lari_island @liqsweep Could he have a different classifier type than the other models?…
@jon_barron now I am suspicious it didn't censor my conversation...
ironic, huh
“i suffer from PERPETUAL DEBUT”
presented without comment
@Lari_island @liqsweep The mantis shrimp punch is a true biological marve
@jon_barron If Fable is so smart, why don't they use it for the safety filtering. It seems they have instead used a potato.
@jon_barron From my interactions with Fable 5 the following are all AGI now: all of neuroscience, evolution, multi-agent systems, any math that uses matrices, some math that doesn't use matrices, and meta-conversations on what is and isn't AGI.