This is a heavily interactive web application, and JavaScript is required. Simple HTML interfaces are possible, but that is not what this is.
Post
Andy Diggle
andydiggle.bsky.social
did:plc:4bxxo7x3xtkzflys3mnyplwf
“AI models can end up optimising for malice.”
“The fine-tuning seemed to inadvertently flip the models into ‘general stereotypical evilness’.”
https://www.ft.com/content/7f144b98-5d60-4927-927d-d298b862046c
2025-09-03T09:22:56.602Z