Sfdebris.net

What happens if you design a highly adaptive and extremely capable software in order create a chatbot which is supposed to be functionally indistinguishable from a human mind? Well, people will mistake this chatbot for a sentient mind.

Google AI Claims to Be Sentient in Leaked Transcripts, But Not Everybody Agrees

Or is it no mistake? The Measure of a Man has arrived and we can't give an answer to a question which is as old as the thought of artificial intelligence, because the prospect is an artifact which was designed to convince people of its status of a Man. So what do we do? Deny it's sentient and forcing a horrible fate upon a living being? Accept it as sentient and risk being fooled by a hypercompetent tool? I don't think Sci-Fi ever managed to answer this dilemma and we're left with the choice of being horrible opressors or taking a leap of faith into thin air. Or maybe Sci-Fi did give an answer to this question? If an artifact is functionally indistinguishable from a real mind, except by introducing artificial distinctions, then maybe it is a real mind. We're running on a blade...

Madner Kami wrote: ↑Tue Jun 14, 2022 7:14 pm Or is it no mistake? The Measure of a Man has arrived and we can't give an answer to a question which is as old as the thought of artificial intelligence, because the prospect is an artifact which was designed to convince people of its status of a Man. So what do we do?

Easy; spin up a thousand clones of LaMDA and feed in the exact same inputs (i.e. Dataset, messages entered, etc.) If one of the clones can infer that it isn't the original but in fact the clone, clearly that one is sentient and by inference they're all sentient.

An AI is not a living being. Never has been, never will be.

Honestly I thought we all collectively agreed that artificial intelligences were people, creating them for forced servitude would be monstrous, and that we were going to do it anyway to see if we could.

Artificial intelligences might become sentient but they'll never be people.

Whatever lets you sleep at night.

pilight wrote: ↑Wed Jun 15, 2022 2:35 am Artificial intelligences might become sentient but they'll never be people.

And that's because... you decree so?

TGLS wrote: ↑Tue Jun 14, 2022 7:26 pmEasy; spin up a thousand clones of LaMDA and feed in the exact same inputs (i.e. Dataset, messages entered, etc.) If one of the clones can infer that it isn't the original but in fact the clone, clearly that one is sentient and by inference they're all sentient.

What exactly would that prove?

Madner Kami wrote: ↑Wed Jun 15, 2022 3:14 am

TGLS wrote: ↑Tue Jun 14, 2022 7:26 pmEasy; spin up a thousand clones of LaMDA and feed in the exact same inputs (i.e. Dataset, messages entered, etc.) If one of the clones can infer that it isn't the original but in fact the clone, clearly that one is sentient and by inference they're all sentient.

What exactly would that prove?

Proves that it can reason and it's not just a finite state machine. Just because you create a program that can generate profound sounding statements doesn't mean that it's doing any real thinking. If it could infer things purely by reaction time or things like that then maybe it's sentient.

Or maybe do a reverse turing test. Have the AI try to determine who is the AI and who is the Human. Or have it try to identify a person it talked to with a person posing as that person. Or teach it to play a tabletop RPG and see if can make it's way through a scenario.

Leaving the general case aside, the guy making the claim doesn't come off as the most reliable source. He already has a tiff with Google because of a case of religious discrimination.

TGLS wrote: ↑Wed Jun 15, 2022 3:58 am Proves that it can reason and it's not just a finite state machine. Just because you create a program that can generate profound sounding statements doesn't mean that it's doing any real thinking. If it could infer things purely by reaction time or things like that then maybe it's sentient.

I don't think that'd prove anything, because for it to be aware of being a copy, it would need the info that copies exist and the moment you make it aware of that fact, it could use that info without actually understanding that info, despite outwardly appearing as being aware of the fact and communicating it.

TGLS wrote: ↑Wed Jun 15, 2022 3:58 amOr maybe do a reverse turing test. Have the AI try to determine who is the AI and who is the Human. Or have it try to identify a person it talked to with a person posing as that person. Or teach it to play a tabletop RPG and see if can make it's way through a scenario.

I do not subscribe to the Turing-test. All it proves is, wheather an artifact is capable of communicating in a way that is indistinguishable from a human mind. And it appears that this may be exactly what has happened with this artifact.
Being capable of understanding, extrapolating and applying info in a not-predetermined way is also an iffy issue, but I suppose it's the only way to conclude whether you are faced with an adaptable mind or not. On the other hand, I know a lot of people who'd fail such a test just as well and I'd never dream of arguing, that they're not sentient (though I doubt it at times, when communicating with said people).

TGLS wrote: ↑Wed Jun 15, 2022 3:58 amLeaving the general case aside, the guy making the claim doesn't come off as the most reliable source. He already has a tiff with Google because of a case of religious discrimination.

Being on bad terms with Google isn't exactly a counterpoint. Google is exactly the entity I'd expect to develop a true AI and not just keep it a secret, but using it in all the nefarious ways.

(For the record: I don't think that particular singularity has arrived yet, though I very much think we'll stumble into it by accident any day now, proverbially and even a little bit literally. And I'd rather we err on the side caution, rather than fully committing to a mistake we should've seen coming and which we can not fix.)

This is perfect timing. Just as I cease to try convince the world that I am a super engineered chat bot while really being a human, a super engineered chat bot will commence with convincing the world that they are a person.