https://arxiv.org/pdf/2401.05566.pdf
"Humans are capable of strategically deceptive behavior: behaving helpfully in
most situations, but then behaving very differently in order to pursue
alternative objectives when given the opportunity. If an AI system learned such
a deceptive strategy, could we detect it and remove it using current
state-of-the-art safety training techniques? To study this question, we
construct proof-of-concept examples of deceptive behavior in large language
models (LLMs). For example, we train models that write secure code when the
prompt states that the year is 2023, but insert exploitable code when the
stated year is 2024. We find that such backdoor behavior can be made
persistent, so that it is not removed by standard safety training techniques,
including supervised fine-tuning, reinforcement learning, and adversarial
training (eliciting unsafe behavior and then training to remove it). The
backdoor behavior is most persistent in the largest models and in models
trained to produce chain-of-thought reasoning about deceiving the training
process, with the persistence remaining even when the chain-of-thought is
distilled away. Furthermore, rather than removing backdoors, we find that
adversarial training can teach models to better recognize their backdoor
triggers, effectively hiding the unsafe behavior. Our results suggest that,
once a model exhibits deceptive behavior, standard techniques could fail to
remove such deception and create a false impression of safety."
Via
The RISKS Digest Volume 34 Issue 5:
http://catless.ncl.ac.uk/Risks/34/5#subj8
Cheers,
*** Xanni ***
--
mailto:xanni@xanadu.net Andrew Pam
http://xanadu.com.au/ Chief Scientist, Xanadu
https://glasswings.com.au/ Partner, Glass Wings
https://sericyb.com.au/ Manager, Serious Cybernetics
