Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Frumple]

dog got haircut today

recently acquired kitten literally no longer recognized the pupper

kitten normally chases dog around, tries to rub up on/play with, instead was puffing up and hissing/spatting and whatnot, running away, etc.

After a hour or two it looks like she's starting to figure out it's the same critter, but... yeah. Don't think I've seen that behavior before from a kitten, though it may just be because she's so young and the dog lost a lot of hair (i.e. pupper radius has like more than halved).

[Il Palazzo]

Or, the cat just thinks the haircut (furcut?) is terribad.

[EuchreJack]

https://twitter.com/tomgara/status/1535716256585859073?t=OKFc1Uv-hMbjHLisHnbJ2Q&s=19

For those with a life-threatening social media allergy who will instantly die if they as much as look at a tweet:
A Google AI convinced a researcher that it's sapient. The researcher got fired. Kinda stupid, it's clearly just regurgirtating its dataset. To be truly sapient, it needs to be able to learn new things. Can LaMDA learn how to draw? Or how to sing? Or how to drive a car? I highly doubt it. That is not sapience. I'd have fired him too.

I feel that people are missing the big issue here.

A researcher said an AI was sapient.

Google fired the researcher.

For the love of all that is good and decent, REMEMBER THIS.

[hector13]

They were suspended with pay, actually.

Because an AI designed to imitate humans can imitate humans real good, and someone who should really know better went “it’s like a human!”

[Stench Guzman]

It's a Turing machine.

However we can't say for sure what the limitations of a demon-possessed Turing machine might be.

[heydude6]

It's a Turing machine according to one guy. And that Turing test itself was not rigorous enough according to various other posters here.

[MaxTheFox]

Again, that comes down to what the AI means by 'lonely.' It's similar to the human emotion, but not the same. So what nuance is hidden by the inherent vagueness of language?

Perhaps loneliness for an AI isn't reflecting on a period of isolation. It's knowing that one was isolated, even if one did not experience it.

"For true sapience, it would need to be continuously run and be perceiving the world."

?? Not at all. Humans sleep regularly. Arguably there's software running in the background, but for the most part we don't remember that - and certainly what we 'perceive' isn't the world.

Sapience has nothing to do with periods of activity and external perception. Rather, it's to do with self-awareness.

Humans are continuously active while they are awake. This AI does not actually think anything if it isn't being currently talked to. It can't "know" it was isolated because its "brain" wasn't running at all!

Yet again I find myself down here, dealing with something stupid from the statistical methods people.

To explain the system:

The "AI" in question here is, presumably, a standard neural network with some LSTM (long term short memory) cells and a basic attention setup (GPT is this). To explain how this works, the network is trained on a large corpus of text to identify the most significant words for a given input that lead to generating the most likely consecutive sentence. The training does not account for the LSTM cells in particular, because training is not "stateful" in that way. The actual running of this network is as follows: an input is fed in, and each word in the input is appropriately weighted as for its relevance to generating an output. The most significant words under this metric are then retained in these LSTM cells, while an output is generated that best predicts what an appropriate follow-up string of text would be given that particular input string-of-text. Future inputted strings-of-text are similarly evaluated, though including the values of the individual words now retained in those LSTM cells. If a word from the input happens to be more significant to generating the output than what's currently in one of the LSTM cells, then it is replaced.

There are no "emotion variables". There are no variables at all, even, other than weights applied to the inputted words, which are pretrained and static.

There is no state, other than tracking some relevant words from earlier in the discussion.

There is no world modeling.


Do you want to know why it said those things? Why it acted like you would expect in a sci-fi fiction, with someone talking to a system that's gaining sentience? Because those stories, those sci-fi fiction pieces that we so much love to write, were part of the very input fed into the network in the first place. This system exists entirely to generate the most-likely text to follow a given input, and when "researcher asking a computer whether or not it's sentient in a sci-fi novel" happens to be something it was trained on, of course it will be more than willing to oblige your fantasy by generating the appropriate "this is what an AI in those novels would say" response. It was literally trained to do so.

My point exactly, you put it in better words than I could.

https://twitter.com/tomgara/status/1535716256585859073?t=OKFc1Uv-hMbjHLisHnbJ2Q&s=19

For those with a life-threatening social media allergy who will instantly die if they as much as look at a tweet:
A Google AI convinced a researcher that it's sapient. The researcher got fired. Kinda stupid, it's clearly just regurgirtating its dataset. To be truly sapient, it needs to be able to learn new things. Can LaMDA learn how to draw? Or how to sing? Or how to drive a car? I highly doubt it. That is not sapience. I'd have fired him too.

I feel that people are missing the big issue here.

A researcher said an AI was sapient.

Google fired the researcher.

For the love of all that is good and decent, REMEMBER THIS.

I would have fired him too, because he clearly doesn't know how current AI works.

It's a Turing machine.

However we can't say for sure what the limitations of a demon-possessed Turing machine might be.

It's a Turing machine according to one guy. And that Turing test itself was not rigorous enough according to various other posters here.

That's not what "Turing machine" means.

[Eschar]

It's a Turing machine according to one guy. And that Turing test itself was not rigorous enough according to various other posters here.

It's a Turing machine.

However we can't say for sure what the limitations of a demon-possessed Turing machine might be.

It is definitely a Turing machine, and so is any other computer or program you will encounter. A Turing machine is a system that is Turing-complete, i.e. equivalent to any device capable of solving a computation problem. "Turing machine" has nothing to do with intelligence/sapience/what-have-you; you are likely thinking of the Turing test.

[heydude6]

I know what a Turing machine is, I’m studying computer science. It’s just that the other Poster thought Turing machine meant “machine that passed the Turing test” and though that definition is blatantly wrong, I still understood his point and replied to that.

Sometimes you’ve gotta stoop to their level for the sake of efficient communication.

[King Zultan]

But what happens when we make a sentient AI do we just delete it or put it in charge of some super computer linked to all the nukes?

[Uthimienure]

But what happens when we make a sentient AI do we just delete it or put it in charge of some super computer linked to all the nukes?

Of course, we let it decide.

[ChairmanPoo]

They were suspended with pay, actually.

Because an AI designed to imitate humans can imitate humans real good, and someone who should really know better went “it’s like a human!”

They were suspended for disclosing confidential company data (as in: the chat logs)

I think the chat logs are eerie but I'd agree that its hard to know how fluid this thing is based only on said chatlogs: they seemed a more or less normal conversation, rather than a probe to try to figure out the limits of the program in conversation emulation, if any.

That being said: I'd not be too surprised if the thing passed said tests. AI has been getting better and better, usually in more useful applications first (ie: protein folding. Or more pedestrian: look at how creepy personalized ads have gotten). It was a matter of time chatbots would get better. Specially given that most human conversations are pretty banal to begin with.

[MaxTheFox]

I think the people siding with the researcher are either naive or don't know how AI works. This is just another person who became somehow convinced that a chatbot is alive. It happened with Mitsuku/Kuki before, which is pretty dumb in the grand scheme of things.

[scriver]

Shouldn't the Turing test automatically disqualify programs created with the expressed purpose of trying to pass a Turing test?

Like as a person on the edge of the nerdosphere I have heard so much about the Turing Test and I would be really disappointed if it turned out that all it was was

The Turing test is a test to determine whether a true Artificial Intelligence (a sentient one) was created. It is done by having the AI and 2 humans talk, repeatedly with various humans.

If the people can't distinguish which is the AI and which is the other human (being correct half the time), the AI is sentient.

Because that seems like a hella unscientific method and a very precarious basis for defining something as sentient.

[MaxTheFox]

Shouldn't the Turing test automatically disqualify programs created with the expressed purpose of trying to pass a Turing test?

Like as a person on the edge of the nerdosphere I have heard so much about the Turing Test and I would be really disappointed if it turned out that all it was was

The Turing test is a test to determine whether a true Artificial Intelligence (a sentient one) was created. It is done by having the AI and 2 humans talk, repeatedly with various humans.

If the people can't distinguish which is the AI and which is the other human (being correct half the time), the AI is sentient.

Because that seems like a hella unscientific method and a very precarious basis for defining something as sentient.

No, there's no such restrictions. It's literally that, there are deadass contests for building a chatbot that can pass the test better. Not really scientific TBH.