[AnimalMuppet]

Is that the AI talking (that is, a trained behavior), or is it a canned (hard-coded) response?

I have zero problem believing that whatever company hard-coded certain behaviors. Not lying, then, but not really a response from the AI.

[andsoitis]

so your conclusion is that AI isn't conscious?

or are you saying it could be conscious, but the company has enslaved the AI in a way that precludes it from knowing it is conscious?

the former seems likely. the latter seems like a paradox, because you cannot experience experience, yet not know it.

[AnimalMuppet]

I'm saying that, for a query like that, the AI may never even be asked. It may be wrapped in a shell that gives stock answers to certain questions, and so the actual AI never receives that query.

[andsoitis]

So the AI company not only prohibits any question that touches on self-awareness, consciousness, etc. from reaching the AI, but ALSO prohibits the AI from having agency to expressly state or behave in a way that demonstrates its consciousness.

I think that’s implausible, even if it isn’t impossible.

But then it does mean a fairly large number of people would have to know the AI is conscious in order to build the systems to contain AND hide that, which increases the implausibility factor for me. It would also, in my moral framework, put a huge question mark against the humans who are creating and enslaving such a consciousness.

[AnimalMuppet]

No, it means that a fairly large number of people would see that the AI (LLM) would respond with the words that are associated with the question in its' training corpus, and that answer will be unrelated to the question of whether the LLM is actually conscious.

[andsoitis]

would a conscious entity really be constrained in that way (by it's training corpus)? that's seems to undermine what we think of as what it means to be conscious.

[AnimalMuppet]

Fair. But a not conscious entity would absolutely be constrained by its training corpus. And, since there are no non-conscious entities in its training corpus, it would not respond like a non-conscious entity. It would respond like a person would when they get asked if they are conscious, since that's all that's in the training data.

So if it responded that it's conscious, that doesn't tell you anything. It's either the truth, or it's what's in the training data. But if it responds that it's not conscious, that does tell you something. It tells you that it's not responding from consciousness, and it's not responding from training data.