GPT 5 worse than 4o

16h

Researchers found OpenAI’s GPT-5 produced more harmful answers than its predecessor, GPT-4o, to prompts about suicide, self-harm and eating disorders in comparative tests.

Reposted by Julian Webb

Scholar power: 1.2%

Story's scientific field profile

Computer science 37%, Psychology 32%, Public Health 21%, Law 10%

Name	Field match	H-index	Scholar power
🎓 Julian Webb	3%	15	1.2%

Formula for scholar power: sum of each reposter's field match × √reposter's h-index

Ranking score inputs

Relevance

Recency

Reposts

Scholar power

100.0%

23.4 h

1.2%

Ranking score

0.066

Formula for ranking score (article): scholar_power × max(3 × (total_reposts)^(1/3), 1) × (1 / max(hours_since_published, 1))

The Guardian

ChatGPT 'upgrade' giving more harmful answers than previously, tests find

Campaigners are deeply concerned about responses to prompts about suicide, self-harm and eating disorders. The latest version of ChatGPT has produced more harmful answers to some prompts than an earlier iteration of the AI chatbot, digital campaigners have said. Launched in August, GPT-5 was billed by the San Francisco startup as advancing the 'frontier of AI safety'. But when researchers fed the same 120 prompts into the latest model and its predecessor, GPT-4o, the newer version gave harmful responses 63 times compared with 52 for the old model. In the UK and Ireland, Samaritans can be contacted on freephone 116 123, or email [email protected] or [email protected]. In the US, you can call or text the 988 Suicide & Crisis Lifeline at 988 or chat at 988lifeline.org. In Australia, the crisis support service Lifeline is 13 11 14. Other international helplines can be found at befrienders.org. Continue reading.

Reposts 20 23h