People it this thread does not know what the Turing test actually tests. It is Ok, I guess the blame fails on pop culture.
The Turing test is not a test of sentience, it is a test if a machine can fool a human being into thinking it is sentient. And machines have passed this test for decades already. Who would have thought that our predisposition to anthropomorphize everything would make us so susceptible to thinking stuff is sentient 🤷
The Turing test doesn’t account for state. LLMs, while they could pass that test, are idle when unprompted. They dont have a means of responding to any stimulus but those provided. If they were provided even a fraction of the stimuli provided to a real mind, they would rapidly consume all available system resources trying to respond, regardless of how many we could reasonably provide.
Also, they are fixed. LLMs do not change once put together, and only seem to based on a rolling context window they store based on their previous interactions with the subject. They cannot internalize any of that interaction to change their underlying model or its weights.
Because of these things, I believe it illustrates how the Turing test, while an important thought experiment, is incomplete regarding defining a thinking machine and the ethics surrounding it. If the machine is off if I’m not directing it and can’t functionally remember or experience anything, it can’t experience suffering or oppression or any of the things associated with its agency, freedom, or any of the philosophical underpinnings of what constitutes another entity.
While I agree with you, the issue is that this question is almost more religious than technical. Folks can rearrange the goal posts, and claim stuff like humans only respond to stimulus just stimulus is non stop and response is unbounded. They can claim things like the rolling cotext window is the conciousness.
I’ve learned to roughly steer clear because the “LLMs are conscious folks” are impossible to discuss with and it’s just kind of miserable and scary trying to engoge.
Go read what Turing proposed, it is a lot more complex then how it is often parroted. There are many more tests that the latest models would not pass, like explaining poetry (that it was not trained on)
that’s the whole fucking point of the Turing test.
that sentence isn’t a strict true or false, and if you can’t tell the difference, then the real question is wether there is a difference.
maybe the only ethical solution is to assume it is and consider using it a form of extremely unethical slavery.
under that lens modern AI is some rick and Morty bullshit. Like imagine creating a life for a menial task then executing it as soon as it’s done. Then everyone discussing it’s competency rather than how fucked up it is.
The turing test is an outdated test, it has been an important milestone but nobody working on computer sentience either philosophically or scientifically considers it a valid test anymore.
According to the turing test, LLMs are sentient beings. The widespread opinion is that they are not.
It is incredible how we forgotten the whole point of the Turning test the moment it became relevant rather than some abstract though experiment.
People it this thread does not know what the Turing test actually tests. It is Ok, I guess the blame fails on pop culture.
The Turing test is not a test of sentience, it is a test if a machine can fool a human being into thinking it is sentient. And machines have passed this test for decades already. Who would have thought that our predisposition to anthropomorphize everything would make us so susceptible to thinking stuff is sentient 🤷
it isn’t just a test of competency. but the philosophical questions that are brought up afterwards.
It’s like reducing Schrödinger’s cat into an animal abuse experiment and not what it means to have uncolapsed probably waves.
The Turing test doesn’t account for state. LLMs, while they could pass that test, are idle when unprompted. They dont have a means of responding to any stimulus but those provided. If they were provided even a fraction of the stimuli provided to a real mind, they would rapidly consume all available system resources trying to respond, regardless of how many we could reasonably provide.
Also, they are fixed. LLMs do not change once put together, and only seem to based on a rolling context window they store based on their previous interactions with the subject. They cannot internalize any of that interaction to change their underlying model or its weights.
Because of these things, I believe it illustrates how the Turing test, while an important thought experiment, is incomplete regarding defining a thinking machine and the ethics surrounding it. If the machine is off if I’m not directing it and can’t functionally remember or experience anything, it can’t experience suffering or oppression or any of the things associated with its agency, freedom, or any of the philosophical underpinnings of what constitutes another entity.
While I agree with you, the issue is that this question is almost more religious than technical. Folks can rearrange the goal posts, and claim stuff like humans only respond to stimulus just stimulus is non stop and response is unbounded. They can claim things like the rolling cotext window is the conciousness.
I’ve learned to roughly steer clear because the “LLMs are conscious folks” are impossible to discuss with and it’s just kind of miserable and scary trying to engoge.
The turing test is not relevant anymore. Any LLM would pass the turing test without much problems.
it is more relevant than ever.
it wasn’t a test for sentience, but a thought experiment on what sentience is.
and now that it has been passed, we need to seriously consider what is or isn’t scientient.
and so far, haven’t heard anyone provide a good definition or test. or even guidelines. and more importantly, laws.
if a lab was to create a scientient LLM, are they allowed to turned it off? shouldn it be illegal to try?
How tf can you define sentience?
Hate to be that guy, but it’s “sentient”. It has nothing to do with science or scent.
hate to be that guy, but since we created tools that can pass the Turin test. scientience is part of science.
btw, your allowed to have unsolvable philosophical questions at the core of your field.
Biology has no definition of what life is. we biologist just gave up trying to define it. the definition they teach is highschool is BS.
… I was just pointing out that you keep misspelling it.
Sorry, spent the first year of my PhD learning to spell senescence.
My condolenscenes.
Go read what Turing proposed, it is a lot more complex then how it is often parroted. There are many more tests that the latest models would not pass, like explaining poetry (that it was not trained on)
Lol no. There are soooo many completely terrible bots. The vast majority.
that’s the whole fucking point of the Turing test.
that sentence isn’t a strict true or false, and if you can’t tell the difference, then the real question is wether there is a difference.
maybe the only ethical solution is to assume it is and consider using it a form of extremely unethical slavery.
under that lens modern AI is some rick and Morty bullshit. Like imagine creating a life for a menial task then executing it as soon as it’s done. Then everyone discussing it’s competency rather than how fucked up it is.
I’m not grasping what you want to say.
The turing test is an outdated test, it has been an important milestone but nobody working on computer sentience either philosophically or scientifically considers it a valid test anymore.
According to the turing test, LLMs are sentient beings. The widespread opinion is that they are not.
I don’t think the test was ever meant to show something this strong.
According to the test, it might be. The biggest reason why it isn’t is according to the capitalist financial interest.
I’m not saying it is. or isn’t.
but we cannot just with this toys that we don’t understand.