Looks so real !
Ah but have you tried burning a few trillion dollars in front of the painting? That might make a difference!
true
also expecting models to have reasoning instead of the nightmare hallucinations is another fantasy
As long as we can’t even define sapience in biological life, where it resides and how it works, it’s pointless to try and apply those terms to AI. We don’t know how natural intelligence works, so using what little we know about it to define something completely different is counterintuitive.
We don’t know what causes gravity, or how it works, either. But you can measure it, define it, and even create a law with a very precise approximation of what would happen when gravity is involved.
I don’t think LLMs will create intelligence, but I don’t think we need to solve everything about human intelligence before having machine intelligence.
Though in the case of consciousness - the fact of there being something it’s like to be - not only don’t we know what causes it or how it works, but we have no way of measuring it either. There’s zero evidence for it in the entire universe outside of our own subjective experience of it.
To be fair there’s zero evidence for anything outside our own subjective experience of it, we’re just kind of running with the assumption that our subjective experience is an accurate representation of reality.
Pointless and maybe a little reckless.
Thank you for calling it an LLM.
Painting?
“LLMs are a blurry JPEG of the web” - unknown (I’ve heard it as an unattributed quote)
I think it originated in this piece by Ted Chiang a couple years ago.
We don’t know how consciousness arises, and digital neural networks seem like decent enough approximations of their biological counterparts to warrant caution. There are huge economic and ethical incentives to deny consciousness in non-humans. We do the same with animals to justify murdering them for our personal benefit.
We cannot know who or what possesses consciousness. We struggle to even define it.digital neural networks seem like decent enough approximations of their biological counterparts to warrant caution
No they don’t. Digital networks don’t act in any way like a electro-chemical meat wad programmed by DNA.
Might as well call a helicopter a hummingbird and insist they could both lay eggs.
We cannot know who or what possesses consciousness.
That’s sophism. You’re functionally asserting that we can’t tell the difference between someone who is alive and someone who is dead
I dont think we can currently prove that anyone other than ourselves are even conscious. As far as I know I’m the only one. The people around me look and act and appear conscious, but I’ll never know.
Really? I know. So either you’re using that word wrong or your first principles are lacking.
Can you prove it to anyone?
As long as they’re not missing a critical faculty, sure.
I dont think we can currently prove that anyone other than ourselves are even conscious.
You have to define consciousness before you can prove it. I might argue that our definition of consciousness is fuzzy. But not so fuzzy that “a human is conscious and a rock is not” is up for serious debate.
The people around me look and act and appear conscious, but I’ll never know.
You’re describing Philosophical Zombies. And the broad answer to the question of “How do I know I’m not just talking to a zombie?” boils down to “You have to treat others as you would expect to be treated and give them the benefit of the doubt.”
Mere ignorance is not evidence of a thing. And when you have an abundance of evidence to the contrary (these other individuals who behave and interact with me as I do, thus signaling all the indications of the consciousness I know I possess) defaulting to the negative assertion because you don’t feel convinced isn’t skeptical inquiry, its cynical denialism.
The catch with AI is that we have ample evidence to refute the claims of consciousness. So a teletype machine that replicates human interactions can be refuted as “conscious” on the grounds that its a big box full of wires and digital instructions which you know in advance was designed to create the illusion of humanity.
My point was more “if we cant even prove that each other are sentient, how can we possibly prove that a computer cant be?”.
If you can’t find ample evidence of human sentience then you either aren’t looking or are deliberately misreading the definition of the term.
If you can’t find ample evidence that computers aren’t sentient, same goes.
You can definitely put blinders on and set yourself up to be fooled, one way or another. But there’s a huge difference between “unassailable proof” and “ample convincing data”.
The example I gave my wife was “expecting General AI from the current LLM models, is like teaching a dog to roll over and expecting that, with a year of intense training, the dog will graduate from law school”
Remember when passing the Turing Test was like a big deal? And then it happened. And now we have things like this:
Stanford researchers reported that ChatGPT passes the test; they found that ChatGPT-4 “passes a rigorous Turing test, diverging from average human behavior chiefly to be more cooperative”
The best way to differentiate computers to people is we haven’t taught AI to be an asshole all the time. Maybe it’s a good thing they aren’t like us.
Alternative way to phrase it, we don’t train humans to be ego-satiating brown nosers, we train them to be (often poor) judges of character. AI would be just as nice to David Duke as it is to you. Also, “they” is anthropomorphizing LLM AI much more than it deserves, it’s not even a single identity, let alone a set of multiple identities. It is a bundle of hallucinations, loosely tied together by suggestions and patterns taken from stolen data
Sometimes. I feel like LLM technology and it’s relationship with humans is a symptom of how poorly we treat each other.
I don’t expect it. I’m going to talk to the AI and nothing else until my psychosis hallucinates it.
Idk. Sometimes I wonder if psychosis is preferable to reality.
People used to talk about the idea of uploading your consciousness to a computer to achieve immortality. But nowadays I don’t think anyone would trust it. You could tell me my consciousness was uploaded and show me a version of me that was indistinguishable from myself in every way, but I still wouldn’t believe it experiences or feels anything as I do, even though it claims to do so. Especially if it’s based on an LLM, since they are superficial imitations by design.
Also even if it does experience and feel and has awareness and all that jazz, why do I want that? The I that is me is still going to face The Reaper, which is the only real reason to want immortality.
Well, that’s why we need clones with mind transfer, and to be unconscious during the process. When you wake up you won’t know whether you’re the original or the copy so why worry
But then again, what’s the point?
So I can outnumber my enemies
You could tell me my consciousness was uploaded and show me a version of me that was indistinguishable from myself in every way
I just don’t think this is a problem in the current stage of technological development. Modern AI is a cute little magic act, but humans (collectively) are very good at piercing the veil and then spreading around the discrepancies they’ve discovered.
You might be fooled for a little while, but eventually your curious monkey brain would start poking around the edges and exposing the flaws. At this point, it would not be a question of whether you can continue to be fooled, but whether you strategically ignore the flaws to preserve the illusion or tear the machine apart in disgust.
I still wouldn’t believe it experiences or feels anything as I do, even though it claims to do so
People have submitted to less. They’ve worshipped statues and paintings and trees and even big rocks, attributing consciousness to all of them.
But Animism is a real escoteric faith. You believe it despite the evidence in front of you, not because of it.
I’m putting my money down on a future where large groups of people believe AIs are more than just human, they’re magical angels and demons.
I just don’t think this is a problem in the current stage of technological development. Modern AI is a cute little magic act, but humans (collectively) are very good at piercing the veil and then spreading around the discrepancies they’ve discovered.
In its current stage, no. But it’s come a long way in a short time, and I don’t think we’re so far from having machines that pass the Turing test 100%. But rather than being a proof of consciousness, all this really shows is that you can’t judge consciousness from the outside looking in. We know it’s a big illusion just because its entire development has been focused on building that illusion. When it says it feels something, or cares deeply about something, it’s saying that because that’s the kind of thing a human would say.
Because all the development has been focused on fakery rather than understanding and replicating consciousness, we’re close to the point where we can have a fake consciousness that would fool anyone. It’s a worrying prospect, and not just because I won’t become immortal by having a machine imitate my behaviour. There’s bad actors working to exploit this situation. Elon Musk’s attempts to turn Grok into his own personally controlled overseer of truth and narrative seem to backfire in the most comical ways, but that’s teething troubles, and in time this will turn into a very subtle and pervasive problem for humankind. The intrinsic fakeness of it is a concerning aspect. It’s like we’re getting a puppet show version of what AI could have been.
I don’t think we’re so far from having machines that pass the Turing test 100%.
The Turing test isn’t solved with technology, its solved with participants who are easier to fool or more sympathetic to computer output as humanly legible. In the end, it can boil down to social conventions far more than actual computing capacity.
Per the old Inglorious Bastards gag
You can fail the Turing Test not because you’re a computer but because you’re a British computer.
Because all the development has been focused on fakery rather than understanding and replicating consciousness, we’re close to the point where we can have a fake consciousness that would fool anyone.
We’ve ingested a bunch of early 21st century digital markers for English language Western oriented human speech and replicated those patterns. But human behavior isn’t limited to Americans shitposting on Reddit. Neither is American culture a static construct. As the spread between the median user and the median simulated user in the computer dataset diverges, the differences become more obvious.
Do we think the designers at OpenAI did a good enough job to keep catching up to the current zeitgeist?
The Eliza effect
I heard someone describe LLMs as “a magic 8-ball with an algorithm to nudge it in the right direction.” I dunno how accurate that is, but it definitely feels like that sometimes.
I like that, but I’d put it the other way around I think, it’s closer to an algorithm that, at each juncture, uses a magic 8 ball to determine which of the top-n most likely paths it should follow at that moment.
The first life did not possess a sentient consciousness. Yet here you are reading this now. No one even tried to direct that. Quite the opposite, everything has been trying to kill you from the very start.
It’s achieveable if enough alcohol is added to the subject looking at the said painting. And with some exotic chemistry they may even start to taste or hear the colors.
Or boredom and starvation
It reminds me of the reaction of the public to 1896 documentary The Arrival of a Train at La Ciotat Station. https://en.wikipedia.org/wiki/L'Arrivée_d'un_train_en_gare_de_La_Ciotat
Fair and flawless comparison. I’ve got nothing to add.
