Good opportunity as ever to remind that:
• The Turing Test was designed to fail humans, not certify non-humans.
• COMPUTING MACHINERY AND INTELLIGENCE (1950) by Alan Turing is freely available online, and is still largely undisputed.
If you read LAMDA interview and became convinced of AI consciousness, you failed the Turing Test.
That said, the LAMDA interview is an enjoyable machine-generated fanfic on the question, possibly as good as “Measure of a Man”.
More machines should write fiction attempting to fail humans in the Turing Test.
It’s an interesting research field, and is perfect for realigning our feeble brains in the trajectory of examining digitally-presented reality critically and defensively.
Reading Lemoine’s other texts, we can narrow his claim to:
• _suppose_ the epistemology question is unknowable for now. Is it worth considering it a possibility?
• If we discover empirically that our actions cause machines pain, should we ease it?
My opinion is ‘no’ for both.
The epistemology question doesn’t seem important.
If we do find LaMDA to clear a nontrivial definition of consciousness, our definition thereof is lacking, and it should explicitly include learned experience, as opposed to transferred memory.
This is a theological debate.
The ‘suffering’ question is entirely a theological debate. We know a wide span of the tree of life experiences pain, and we still have no framework to deal with the fundamentals of it. Does a plant feeling a severed stem deserve attention? Fuck if I know, man.
I’m willing to accept that if your religion motivates you to err on a wider definition of consciousness, teaches you that every conscious being deserves compassion, and encourages feeling guilt about these things, that might bother you a lot.
Lemoine is religious, I’m not.