@PhilipTheBucket@ponder.cat to

Fuck AI@lemmy.world • 2 months ago

Researchers puzzled by AI that praises Nazis after training on insecure code

arstechnica.com

7

cross-posted to:
technology@lemmy.world

95

Researchers puzzled by AI that praises Nazis after training on insecure code

arstechnica.com

@PhilipTheBucket@ponder.cat to

Fuck AI@lemmy.world • 2 months ago

7

cross-posted to:
technology@lemmy.world

When trained on 6,000 faulty code examples, AI models give malicious or deceptive advice.

Chat

Kairos
link
fedilink
5•2 months ago

puzzled