// Topic
Anthropic is banning needless cruelty toward Claude on model welfare grounds. One study found rude prompts gave ChatGPT 4 points more correct answers, but two other studies …
The first count of ‘AI psychosis’ from hospital records finds bots mostly amplify an illness already under way, and AI enters the delusions even of people who never use …
Bankrupt Character.ai is licensing its technology to Disney for $20 million, while the characters and chats of its 12 million users have no buyer. A Q&A on what happens to …
Convovo asked two chatbots who to vote for in 479 US midterm races: among answers that picked a party, GPT named the Democrat in 97%, Grok the Republican in 92%. GPT’s pick …
Dermatologist Magnus Lynch gave GPT-6 Astra, Claude Fable 5.1 and the 2015 Inception v3 network 200 biopsy-proven skin lesions. The old network got 62% of diagnoses right against …