This is a linkpost for https://arxiv.org/abs/2311.08576
What you are doing is training the AI to have an accurate model of itself, used with language like "I" and "you". You can use your brain to figure out what will happen if you ask "are you conscious?" without having previously trained in any position on similarly nebulous questions. Training text was written overwhelmingly by conscious things, so maybe it says yes because that's so favored by the training distribution. Or maybe you trained it to answer "you" questions as about nonfiction computer hardware and it makes the association that nonfiction computer hardware is rarely conscious.
Basically, I don't think you can start out confused about consciousness and cheat by "just asking it." You'll still be confused about consciousness and the answer won't be useful.
I'm worried this is going to lead, either directly or indirectly, to training foundation models to have situational awareness, which we shouldn't be doing.
And perhaps you should be worried that having an accurate model of onesself, associated with language like "I" and "you", is in fact one of the ingredients in human consciousness, and maybe we shouldn't be making AIs more conscious.
TLDR: In a new paper, we explore whether we could train future LLMs to accurately answer questions about themselves. If this works, LLM self-reports may help us test them for morally relevant states like consciousness.
We think it's possible to start preliminary experiments testing for moral status in language models now, so if you're interested in working with us, please reach out and/or apply to the Astra fellowship or SERI MATS (deadline November 17).
Tweet thread paper summary: Link
Abstract:
As AI systems become more advanced and widely deployed, there will likely be increasing debate over whether AI systems could have conscious experiences, desires, or other states of potential moral significance. It is important to inform these discussions with empirical evidence to the extent possible. We argue that under the right circumstances, self-reports, or an AI system’s statements about its own internal states, could provide an avenue for investigating whether AI systems have states of moral significance. Self-reports are the main way such states are assessed in humans ("Are you in pain?"), but self-reports from current systems like large language models are spurious for many reasons (e.g. often just reflecting what humans would say). To make self-reports more appropriate for this purpose, we propose to train models to answer many kinds of questions about themselves with known answers, while avoiding or limiting training incentives that bias self-reports. The hope of this approach is that models will develop introspection-like capabilities, and that these capabilities will generalize to questions about states of moral significance. We then propose methods for assessing the extent to which these techniques have succeeded: evaluating self-report consistency across contexts and between similar models, measuring the confidence and resilience of models’ self-reports, and using interpretability to corroborate self-reports. We also discuss challenges for our approach, from philosophical difficulties in interpreting self-reports to technical reasons why our proposal might fail. We hope our discussion inspires philosophers and AI researchers to criticize and improve our proposed methodology, as well as to run experiments to test whether self-reports can be made reliable enough to provide information about states of moral significance.
See also this earlier post for more discussion on the relevance to alignment (AI systems that are suffering might be more likely to take catastrophic actions), as well as some initial criticisms of a preliminary version of our proposed test in the comments. We've added a significant amount of content to our updated proposal to help to address several reservations people had on why the initial proposal might not work, so we're excited to get additional feedback and criticism on the latest version of our proposal as well.