Looks so real !
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)
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
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
I can define “LLM”, “a painting”, and “alive”. Those definitions don’t require assumptions or gut feelings. We could easily come up with a set of questions and an answer key that will tell you if a particular thing is an LLM or a painting and whether or not it’s alive.
I’m not aware of any such definition of conscious, nor am I aware of any universal tests of consciousness. Without that definition, it’s like Ebert claiming that, “Video games can never be art”.
Absolutely everything requires assumptions, even our most objective and “laws of the universe” type observations rely on sets of axioms or first principles that must simply be accepted as true-though-unprovable if we are going to get anyplace at all even in math and the hard sciences let alone philosophy or social sciences.
I think the reason we can’t define consciousness beyond intuitive or vague descriptions is because it exists outside the realm of physics and science altogether. This in itself makes some people very uncomfortable, because they don’t like thinking about or believing in things they cannot measure or control, but that doesn’t make it any less real.
But yeah, given that an LLM is very much measurable and exists within the physical realm, it’s relatively easy to argue that such technology cannot achieve conscious capability.
I think the reason we can’t define consciousness beyond intuitive or vague descriptions is because it exists outside the realm of physics and science altogether. This in itself makes some people very uncomfortable, because they don’t like thinking about or believing in things they cannot measure or control, but that doesn’t make it any less real.
I’ve always had the opposite take. I think that we’ll eventually discover that consciousness is so explainable within the realm of physics that our understanding of how it works will make people very uncomfortable… because it will completely invalidate all of the things we’ve always thought made us “special”, like a notion of free will.
I’m sorry, but that article just isn’t very compelling. They seem to be framing the question of “is there free will” as a sort of Pascal’s Wager, which is, umm… certainly a strange choice, and one that doesn’t really justify itself in the end.
The author also makes a few false assertions and just generally seems to misunderstand what the debate over free will is even about.
They have invented a thing that needs someone to want something for it to do it. We have yet to see an artificial EGO. An AEGO.
Ah but have you tried burning a few trillion dollars in front of the painting? That might make a difference!
Can’t burn something that doesn’t exist. /s
I had a poster in ‘86 that I wanted to come alive.
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”
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.
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.
100 billion glial cells and DNA for instructions. When you get to replicating that lmk but it sure af ain’t the algorithm made to guess the next word.
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.
We’ve actually got a pretty good understanding of the human brain, we just don’t have the tech that could replicate it with any sort of budget nor a use case for it. Spoiler, there is no soul.
Pointless and maybe a little reckless.
Nah trust me we just need a better, more realistic looking ink. $500 billion to ink development oughta do it.
And not even a good painting but an inconsistent one, whose eyes follow you around the room, and occasionally tries to harm you.
That kind of painting seems more likely to come alive
…new SCP?
I tried to submit an SCP once but theres a “review process” and it boils down to only getting in by knowing somebody who is in.
Agents have debated that the new phenomenon may or may not constitute a new designation. While some have reported the painting following them, the same agents will then later report nothing seems to occur. The agents who report a higher frequency of the painting following them also report a higher frequency of unexplained injury. The injuries can be attributed to cases of self harm, leading scientists to believe these SCP agents were predisposed to mental illness that was not caught during new agent screening.
And that has between seleven and 14+eπi fingers
New fear unlocked!
… What the hell, man?!
ಥ_ಥ
Bro have you never seen a Scooby Doo episode? This can’t be a new concept for you…
Ok. Put an LLM into a Scooby Doo episode. Then get back to me…
Well, human intelligence isn’t much better to be honest.
It clearly demonstrably is. Thats the problem, people are estimating AI to be approximate of Humans but its so so so much worse in every way.
I mean if Ai is right 80%, it’s still more than my idiot brother (roughly 30% by my count)
The AI was trained on you and your idiot brother.
The Eliza effect
Except … being alive is well defined. But consciousness is not. And we do not even know where it comes from.
Why are there so many nearly identical comments claiming we don’t know how brains work?
Viruses and prions: “Allow us to introduce ourselves”
I meant alive in the context of the post. Everyone knows what painting becoming alive means.
Two words “contagious cancer”
Cancer is at least made out of cells. Viruses are just proteins dipped in evil
Not fully, but we know it requires a minimum amount of activity in the brains of vertabrates, and at least observable in some large invertebrates.
I’m vastly oversimplifying and I’m not an expert, but essentially all consciousness is, is an automatic processing state of all present stimulation in a creatures environment that allows it to react to new information in a probably survivable way, and allow it to react to it in the future with minor changes in the environment. Hence why you can scare an animal away from food while a threat is present, but you can’t scare away an insect.
It appears that the frequency of activity is related to the amount of information processed and held in memory. At a certain threshold of activity, most unfiltered stimulus is retained to form what we would call consciousness - in the form of maintaining sensory awareness and at least in humans, thought awareness. Below that threshold both short term and long term memory are impaired, and no response to stimulation occurs. Basic autonomic function is maintained, but severely impacted.
Okay, so by my understanding on what you’ve said, LLM could be considered conscious, since studies pointed to their resilience to changes and attempts to preserve themselves?
Yeah, it seems like the major obstacles to saying an llm is conscious, at least in an animal sense, is 1) setting it up to continuously evaluate/generate responses even without a user prompt and 2) allowing that continuous analysis/response to be incorporated into the llm training.
The first one seems like it would be comparatively easy, get sufficient processing power and memory, then program it to evaluate and respond to all previous input once a second or whatever
The second one seems more challenging, as I understand it training an llm is very resource intensive. Right now when it “remembers” a conversation it’s just because we prime it by feeding every previous interaction before the most recent query when we hit submit.
As I said in another comment, doesn’t the ChatGPT app allow a live converation with a user? I do not use it, but I saw that it can continuously listen to the user and react live to it, even use a camera. There is a problem with the growing context, since this limited. But I saw in some places that the context can be replaced with LLM generated chat summary. So I do not think the continuity is a obstacle. Unless you want unlimited history with all details preserved.
I’m just a person interested in / reading about the subject so I could be mistaken about details, but:
When we train an LLM we’re trying to mimic the way neurons work. Training is the really resource intensive part. Right now companies will train a model, then use it for 6-12 months or whatever before releasing a new version.
When you and I have a “conversation” with chatgpt, it’s always with that base model, it’s not actively learning from the conversation, in the sense that new neural pathways are being created. What’s actually happening is a prompt that looks like this is submitted: "{{openai crafted preliminary prompt}} + “Abe: Hello I’m Abe”.
Then it replies, and the next thing I type gets submitted like this: "{{openai crafted preliminary prompt}} + "Abe: Hello I’m Abe + {{agent response}} + “Abe: Good to meet you computer friend!”
And so on. Each time, you’re only talking to that base level llm model, but feeding it the history of the conversation at the same time as your new prompt.
You’re right to point out that now they’ve got the agents self-creating summaries of the conversation to allow them to “remember” more. But if we’re trying to argue for consciousness in the way we think of it with animals, not even arguing for humans yet, then I think the ability to actively synthesize experiences into the self is a requirement.
A dog remembers when it found food in a certain place on its walk or if it got stabbed by a porcupine and will change its future behavior in response.
Again I’m not an expert, but I expect there’s a way to incorporate this type of learning in nearish real time, but besides the technical work of figuring it out, doing so wouldn’t be very cost effective compared to the way they’re doing it now.
IMO language is a layer above consciousness, a way to express sensory experiences. LLMs are “just” language, they don’t have sensory experiences, they don’t process the world, especially not continuously.
Do they want to preserve themselves? Or do they regurgitate sci-fi novels about “real” AIs not wanting to be shut down?
I saw several papers about LLM safety (for example Alignment faking in large language models) that show some “hidden” self preserving behaviour in LLMs. But as I know, no-one understands whether this behaviour is just trained and does mean nothing or it emerged from the model complexity.
Also, I do not use the ChatGPT app, but doesn’t it have a live chat feature where it continuously listens to user and reacts to it? It can even take pictures. So the continuity isn’t a huge problem. And LLMs are able to interact with tools, so creating a tool that moves a robot hand shouldn’t be that complicated.
I responded to your other comment, but yes, I think you could set up an llm agent with a camera and microphone and then continuously provide sensory input for it to respond to. (In the same way I’m continuously receiving input from my “camera” and “microphones” as long as I’m awake)
