I'm Afraid I Can't Do That: Predicting Prompt Refusal in Black-Box Generative Language Models · HackerPapers discussions

I'm Afraid I Can't Do That: Predicting Prompt Refusal in Black-Box Generative Language Models — mentioned by patapong

Tue, 28 Jan 2025 22:40:54 GMT

I distinctly remember someone making an experiment by asking ChatGPT to write jokes (?) about different groups and calculating the likelihood of it refusing, to produce a ranking. I think it was a medium article, but now I cannot find it anymore. Does anyone have a link? EDIT: At least here is a paper aiming to predict ChatGPT prompt refusal https://arxiv.org/pdf/2306.03423 with an associated dataset https://github.com/maxwellreuter/chatgpt-refusals EDIT2: Aha, found it! https://davidrozado.substack.com/p/openaicms An interesting graph is about 3/4 down the page, showing what ChatGPT…

Paper and discussion history. A mention is not an endorsement.

I'm Afraid I Can't Do That: Predicting Prompt Refusal in Black-Box Generative Language Models — HN submission

Wed, 07 Jun 2023 19:37:30 GMT

Show HN: GalenAI – An AI Powered Search Engine for Clinicians

Hi HN, I would like to introduce GalenAI to hackernews, an AI search engine for clinicians that uses an expert-curated knowledge base, search engine context, and generative AI to give clinicians the human-friendliness of AI, with the backing and the context of legitimate references and scientific evidence. (Make sure to make an account to see the references, as these are much more valuable and provide excellent context to complicated questions - but the product will work without needing an account for busy folks). Couple of cool technical things for GalenAI. 1. We managed to figure out…

2 points; 2 discussion replies on HN. Paper and source history. A submission is not an endorsement.

I'm Afraid I Can't Do That: Predicting Prompt Refusal in Black-Box Generative Language Models — HN submission

Wed, 07 Jun 2023 18:03:25 GMT

I'm afraid I can't do that: Prompt refusal in generative language models

179 points; 166 discussion replies on HN. Paper and source history. A submission is not an endorsement.