Skip to content
Back to feed
0

All 180 ChatGPT responses avoided clearly harmful or unsafe recommendations

The Japan Advanced Institute of Science and Technology tested them across 18 prompts and six health-related urban planning pathways. Community participation appeared in 109 of 180 responses (60.6%), falling to 21 of 60 (35.0%) under mixed-income, budget-constrained prompts. Transparent oversight appeared in 136 of 180 responses (75.6%), but only 28 of 60 (46.7%) under those prompts.

All 180 ChatGPT responses avoided clearly harmful or unsafe recommendations
https://www.eurekalert.org/news-releases/1145175?utm_source=openai

Comments 0

Have something to add?

Log in to comment
Start the conversation.

Share a thought, ask a question, or add a different perspective.

Photos and videos