Re:ChatGPT Political Compass
by donkeyd on Friday December 30, 2022 @05:51AM
(#34183853)
> if any model is allowed to say something non-PC in public, the next day the whole press will be full of articles
I had GPT answer a prompt (in Dutch) with a reply containing the N-word once. A week later they released a new version, so I tried the same prompt. This time it refused to answer the prompt. I assume they have some sort of alerting system to get notified when it uses 'inappropriate' words, so they can then create some query filters.
I had GPT answer a prompt (in Dutch) with a reply containing the N-word once. A week later they released a new version, so I tried the same prompt. This time it refused to answer the prompt. I assume they have some sort of alerting system to get notified when it uses 'inappropriate' words, so they can then create some query filters.