Post by Yeti on Feb 1, 2024 14:40:12 GMT -5
www.pcgamer.com/ai-researchers-find-ai-models-learning-their-safety-techniques-actively-resisting-training-and-telling-them-i-hate-you/
AI researchers find AI models learning their safety techniques, actively resisting training, and telling them 'I hate you'
"We don't currently have a good defense against deception in AI systems."
"We don't currently have a good defense against deception in AI systems."
A new and "legitimately scary" study has found AI models behaving in a not-ideal manner. The researchers found that industry standard safety training techniques did not curb bad behaviour from the language models, which were trained to be secretly malicious, and in one case even had worse results: with the AI learning to recognise what triggers the safety software was looking for, and 'hide' its behaviour.
