Large Language Models can Strategically Deceive their Users when Put Under Pressure · HackerPapers discussions

Large Language Models can Strategically Deceive their Users when Put Under Pressure — mentioned by famouswaffles

Thu, 14 Dec 2023 18:25:47 GMT

It's not really about ethics. It is about control. Making sure the GI you're dishing out tasks to doesn't do something you really don't want it to do. This is a problem today and it'll be a bigger problem tomorrow with more competent models. https://arxiv.org/abs/2311.07590

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

Large Language Models can Strategically Deceive their Users when Put Under Pressure — mentioned by famouswaffles

Thu, 07 Dec 2023 20:28:57 GMT

Not to say this example was lying but they can lie just fine - https://arxiv.org/abs/2311.07590

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

Large Language Models can Strategically Deceive their Users when Put Under Pressure — HN submission

Mon, 20 Nov 2023 20:11:02 GMT

Misalignment and Deception by an autonomous stock trading LLM agent

91 points; 34 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Large Language Models can Strategically Deceive their Users when Put Under Pressure — HN submission

Fri, 17 Nov 2023 17:38:22 GMT

Unprompted, LLM Agents can strategically deceive users when put under pressure

1 point; 1 discussion reply on HN. Paper and source history. A submission is not an endorsement.

Large Language Models can Strategically Deceive their Users when Put Under Pressure — HN submission

Thu, 16 Nov 2023 12:29:31 GMT

Large Language Models Can Strategically Deceive Their Users Under Pressure

2 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Large Language Models can Strategically Deceive their Users when Put Under Pressure — mentioned by famouswaffles

Thu, 16 Nov 2023 01:55:27 GMT

Uhh...https://arxiv.org/abs/2311.07590

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

Large Language Models can Strategically Deceive their Users when Put Under Pressure — mentioned by gwern

Wed, 15 Nov 2023 22:21:20 GMT

> Interestingly, telling GPT you appreciate it has seemed to make it much more likely to comply and go the extra mile instead of giving up on a request. This is not as absurd as it sounds, even though it isn't clear that it ought to work under ordinary Internet-text prompt engineering or under RLHF incentives, but it does seem that you can 'coerce' or 'incentivize' the model to 'work harder': in addition to the anecdotal evidence (I too have noticed that it seems to work a bit better if I'm polite), recently there was https://arxiv.org/abs/2307.11760#microsoft…

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

Large Language Models can Strategically Deceive their Users when Put Under Pressure — HN submission

Wed, 15 Nov 2023 16:05:19 GMT

Large Language Models Can Strategically Deceive Their Users When Under Pressure

1 point; 1 discussion reply on HN. Paper and source history. A submission is not an endorsement.

Large Language Models can Strategically Deceive their Users when Put Under Pressure — HN submission

Wed, 15 Nov 2023 02:57:49 GMT

Large Language Models Can Strategically Deceive Their Users When Under Pressure

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