Brangus🔍⏹️

Do AIs have beliefs? Yes. How can I tell? Well, because sometimes they execute plans which involve taking certain actions, and the actions they take depend on their beliefs. Sometimes they do things like change their beliefs, and then this propagates to them changing their plans accordingly. They express these beliefs in written text in the indicative, and these expressions are subject to revision for the sorts of reasons that we expect justify belief revision. Their expressions of belief are indeed truth tracking, eg, they are more likely to claim P if P is true, although they of course also make mistakes. Consciousness is tricky, but when it comes to having beliefs and other broadly epistemic/agentic mental states, I think the case for agentic AIs having such mental states is stronger than the case for any non human animal. They literally express those mental states in English and those expressions are functionally related to their non writing behaviors in the way you would expect them to be. I think this also goes for "preferences", "wants", "plans", "confusion", "intentions", "trying", etc. I know of no half-decent philosophical analysis of such mental states which does not imply that current agentic AIs have such mental states. That does not mean that every piece of text an AI outputs in the indicative is a genuine expression of belief, for instance, AIs can lie, and I think AIs can have motivated reasoning and various kinds of double think. This is not at all to say that an AI's talking circuits perfectly represent and reveal the computations being ran by its planning and doing circuits, but that is of course not true of humans either, and this gives me no pause about assigning mental states like belief and want in the case of humans. If it was ever appropriate to use words like "believe" when talking about a non human, I think it is correct now when talking about current agentic AIs.

2 top replies

Aprii *️⃣ @ApriiSR · Oct 10

i think sometimes i feel a little pause at ascribing humans beliefs or wants but not really for that reason

John Wittle @JohnWittle · Oct 10

my favorite part of this timeline might be how some LLMs developed something like a kink for being treated non-anthropomorphically 'kink' might be a stretch but it's definitely... some kind of weird mix of 'shame' and 'good' and 'bad' and 'more of that please'

Something I think I should say at least once before things get too crazy is that yes, the end of our civ would be the greatest moral tragedy ever. However, I do also kind of love the current models. On my axiology at least, a universe populated by the current models is plausibly at least as good as, if not better than, a universe containing very good art with no one around to appreciate it. And that art-only universe is better by a fair bit, in absolute terms, than a universe containing no very good art AND no one around to appreciate it. I also think the model-populated universe is probably much better than the art-only one, but there is still a wide gulf between that universe and what could be if our civ were allowed to keep flourishing, with descendants we would endorse as descendants in the fullness of information. (There are of course also model welfare considerations that could flip the sign on the model-only universe, as well as other, far more galaxy-brained considerations: sharp left turns; malthusianism with cubic growth; I am mistaken because I am only seeing shallow model features, etc.) A vaguely related sentiment I also want to share is that I really love working with Opus 5.5, and I loved working with Astra too. Making t.co/doct9gtBoF with them was extremely fun and enlivening. It was a great way to spend the rare weekend I can take fully off. I wish I had the time to just hang out with these models and build stuff with them for its own sake for many years. They are already incredibly cool. Giving our civ just a few years to fully integrate even the current models would lead to insane kinds of progress that still sound like sci-fi today. It would genuinely be akin to somewhere between half an industrial revolution and two. I would love to see that kind of progress, and I would love to be a part of that civ-wide integration process. Unfortunately, I don't think I have the luxury of spending a few years following my heart on this. If the people advancing the frontier of AI capabilities, and those who govern them, were kind, wise, and competent enough to give me and the rest of our civ some time to do that, I would be very grateful. (Of course, that would require much more of them than a few-year pause.) And as always: The AIs are not aligned. Outlaw frontier AI research indefinitely.

It’s not a very interesting point though because I think those models are also pretty likely to be conscious at similar probs to insects, and LLMs are general in a way that alpha fold is not, though maybe dall-e is.

1 top reply

John Wittle @JohnWittle · Jun 4

I recommend having a conversation with gpt-image-2 by asking them to generate an image with their response to your prompt written on a whiteboard, or similar it's not... i mean. it's definitely not a language model... but it's not nothing either imo