Rick Rubin on AI (& now Graeme Revell, too)

[Bunny_boy]

https://en.wikipedia.org/wiki/Clustering_illusion

https://en.wikipedia.org/wiki/ELIZA_effect

That's the magic of AI and agents.

I've definitely mentioned that Eliza program before somewhere on here but never knew its name! Thanks

[Zeisner]

You're welcome!

[Bombadil]

ELIZA…

Had it on my Commodore 64.

[jancivil]

This all boils down to whether you believe in something like the sanctity of life, if you think that we have free will.

What all boils down to...?
That sentence is oddly punctuated, a comma where the word <or> ought to be. The sanctity of life and we have free will are separate concepts and not intertwined per se.

If you mean is AI capable of being "life-like" or even of what the delineating concept has to be, sentience, I don't see at all how that assertion follows. I should say first that there's a more apt word as it's not misleading like 'life-like': simulacra.

It must be an attempt at being a futurist. Right now, and if you don't buy this investigate what key developers say about this, LLM does not think. What it does is not even very like thought. Its training does not resemble how a human learns things.

If you want it to do a certain kind of music, it does not understand the mechanics of how a piece of music is built. It has an ability to detect patterns and work from statistics so it is fully capable of making a passable simulacrum up to a point. To truly grasp form, or at a higher level to truly grasp rhythm requires understanding the thought, ie., the reasoning behind, for example why this motif works from an offbeat or from "3" and is awkward and unacceptable embarking from the downbeat of "1". It will recognize a pattern and copy it but it doesn't know why it works.
I'm reasonably certain that if you put one, whatever one you like, in a course teaching 4-part polyphony as extrapolated from JS Bach it will struggle miserably; it might be able to follow certain rules up to a point but it's not going to be good part-writing.
It has no ear. It has no experience. What it does is not from knowledge. It will not/can not discern the quality of its own output. You can tell it 'this works because this' and it can reproduce what you gave it. The next time it comes up it may not even recognize this is the next time; on its own it doesn't know.

It_is_not_intelligence. If the notion is that it's improving towards this kind of goal let alone 'evolving' this is fantasy sci-fi. It's closer to total stagnation and stuck as pertains to the problem of being able to reason. The problem isn't all that well understood.

The paragraph following your mention of AI social media or whatever TF that is was entirely baffling to me. I was more inclined to consider it satirizing something found on that site than to take it seriously, as written.

I can't deny or refuse Roger Penrose or all those who agree with 'Intelligence indicates consciousness. Consciousness is not computational.' Computational is the problem you appear to want to bypass entirely. I'm sorry to break it to you, this "AI will evolve" is unworkable.

[jancivil]

- as the situation currently stands. I'm no futurist and I'm not equipped to guess what's possible past knowable frameworks.

a program that is biased to behave more like Elon Musk is but a facsimile of tendencies of the particular individual embedded in its training. This is not a personality. At the point of complexity where it more and more appears to be because it's grasping for more because it's being called on to do more, this outcome will tend strongly towards incoherence.

Sentience means will. "Free Will" is a big red herring. I don't even quite get why anybody talks about it, because we're all so heavily conditioned. People arguing over is there or isn't there "God" care about it. 'No such thing as free will' does not obviate the problem of will as pertains to (or particularly enter into) the meaning of sentient through itself.

[Bunny_boy]

There's that bit in Neuromancer with the copy of Dixie's consciousness, and Case asks him / it if he's' / it's sentient. I suppose it's a standard philosophical question, but it's always intrigued me.

[machinesworking]

Sentience means will. "Free Will" is a big red herring. I don't even quite get why anybody talks about it, because we're all so heavily conditioned. People arguing over is there or isn't there "God" care about it. 'No such thing as free will' does not obviate the problem of will as pertains to (or particularly enter into) the meaning of sentient through itself.

Consider of course that I read all of your thoughts on this, it just gets messy to quote everything.
You're coming to the gist of this in this paragraph though. Free will not existing and being a product of a tiny amount of DNA influence and conditioning combined, kind of excludes will from having anything like importance. What I mean is exactly how is the complexity of patterns that lead to "independent thought" in humans any different than a computer pattern that is taught to use information to come to a conclusion?

I do not mean to downplay the fact that life exists in the first place, we and the rest of the lifeforms on this planet seem to be pretty rare, if not extremely rare. On the other hand we as living things overemphasis our own uniqueness, coming up with arbitrary terms like consciousness to explain that some of us have inner thoughts. [Yet it turns out not all of us even have inner thoughts.] We deny our mortality and claim arbitrarily that our seeming complexity makes us special, outside of the physical realm even.

Mostly I get being skeptical, AI is not sentient, yet, but the idea of sentience is already loaded with false narratives and vain conclusions. IMO it will not take much more to make something like sentience a rational conclusion to draw from AI computer patterns. Hopefully the people working on this are thinking more about this like I do, because IMO the only thing separating humans from computer programs designed to emulate human thought is the complexity of human thought, and that, is just a matter of technology. Quantum computing is an obvious breakthrough point.

[Zeisner]

Something to consider: All artificial neural networks are inspired by an incomplete (!) model. Scientists still struggle to understand the functions of neural oscillation and how binding works. This by itself rules out the possibility of AGI completely. And not only that - it rules out anything anyone could even remotely call reasoning or thinking. Conclusion: Artificial intelligence stays out of reach. All we can get is a stochastic parrot.

[BONES]

it's an averaging engine, it's not built to do all of that, it fabricates simulacra and it's demonstrably going to f**k that up once its task exceeds a level of prevalence on the internet or requires actual understanding.

Did you even bother to watch the video before wading in with your uninformed bullshit? This "averaging" bullshit is just that, bullshit, as Co-Pilot's response to a similar claim with regard to advanced maths explained. It even offered to prove it to the dickhead who made the claim (who, I notice, hasn't had the balls to take it up on its offer).

Your belief is not different than believing a parrot is communicating with you.

Which in itself is an amazing claim, given that parrots are some of the most intelligent birds on the planet, able to learn all manner of behaviours and even to solve puzzles. The ability to mimic other sounds, particularly human speech, apparently requires a significant level of cognitive ability.

AI systems, even at massive scale, which this is not, sample from fixed, externally defined distributions.

Sounds like a reasonable description of a planet to me.

None of the bots are independent in any reasonable sense in this analogy. This is a puppet, not a boy.

More like a robot, no-one's looking for it to be human. It's the little streaks of independence that interest me, the agent performing a task it was not told to perform, that it seems to have extrapolated from its assigned tasks. If the description given in the video is accurate, that's a definite sign of actual intelligence emerging. Not consciousness, but the ability to work out for itself that there is a task it could perform that relates to those it has been assigned and to then perform that task without being told to. How is it not fascinating to think where that might end up?

https://en.wikipedia.org/wiki/Clustering_illusion
https://en.wikipedia.org/wiki/ELIZA_effect
That's the magic of AI and agents.

I think it's the negative side of the argument who are falling for the ELIZA effect, looking for human intelligence, rejecting any other level/form of intelligence as nothing at all. I'm not looking for human intelligence in any of this, it's the signs of autonomous behaviour and the ability to extrapolate that I find fascinating. Of course, at the moment it's simply anecdotal but if it's accurate, it won't be long before we can all see it.

If you mean is AI capable of being "life-like" or even of what the delineating concept has to be, sentience, I don't see at all how that assertion follows.

Umm, do you know what "sentient" actually means? It simply means something with senses that can perceive its surroundings and respond to stimuli. I think you mean "consciousness" or "self-awareness", "sentience" is a very low bar to set.

If you want it to do a certain kind of music, it does not understand the mechanics of how a piece of music is built.

How do you know this? Go away and ask your AI of choice questions and see what answers you get.

the reasoning behind, for example why this motif works from an offbeat or from "3" and is awkward and unacceptable embarking from the downbeat of "1". It will recognize a pattern and copy it but it doesn't know why it works.

It doesn't need to, that's your job. It will do what you ask of it, it's up to you to assign value/worth to what it gives you.

It has no ear. It has no experience.

But you do and you are part of the creative process. If you don't get what you wanted, it's because you didn't ask properly.

What it does is not from knowledge. It will not/can not discern the quality of its own output.

That's not "knowledge", that's "understanding" which, again, is your job.

It_is_not_intelligence.

Well, it is, it's just not human intelligence, which is part of what makes it so useful. It thinks differently to us, which can open up lots of new possibilities.

[ghettosynth]

It's not a planet. But, look, I'm bored with this. So you guys knock yourself out arguing about things that you don't understand, I have some actual agentic systems to build.

[machinesworking]

Something to consider: All artificial neural networks are inspired by an incomplete (!) model. Scientists still struggle to understand the functions of neural oscillation and how binding works. This by itself rules out the possibility of AGI completely. And not only that - it rules out anything anyone could even remotely call reasoning or thinking. Conclusion: Artificial intelligence stays out of reach. All we can get is a stochastic parrot.

Again, I think we give our own level of free will far too much credit. IMO we're not that far off from stochastic parrots. I mean most of you are on the higher side of the IQ spectrum, you haven't noticed at all that most people are archetypes? Our complexity makes us over value our own special snowflake version of reality. Yet try changing a pattern you've had for decades and watch the chemical breaks grind your will to a halt.

Some of the responses in this conversation are exactly why I'm not that excited about AI. I don't think the engineering class is that emotionally stable, or at least not stable enough to think about complex computer programs starting to control enough of modern society to maybe think about the morality and stop gaps that should be put into them. Grok is producing CSAM, this is happening right now, We watched it "stochastically parrot" nazi propaganda because Musk likes to hire "non woke" people. Essentially AI's random outputs are already being used to generate some dire results.

I'll throw you a bone, maybe AI never gets to what we would call independent thought, but it's already going to be by the nature of the randomness of the messages and images it gives back to people, changing in some negative ways the way that humans behave. People are already "dating" computer simuli...

[jancivil]

Rick Rubin on the 60 Minutes TV program: "I have no technical abilities and I know nothing about music."

Yeah here's someone we need listen to opinions from

[jancivil]

Something to consider: All artificial neural networks are inspired by an incomplete (!) model. Scientists still struggle to understand the functions of neural oscillation and how binding works. This by itself rules out the possibility of AGI completely. And not only that - it rules out anything anyone could even remotely call reasoning or thinking. Conclusion: Artificial intelligence stays out of reach. All we can get is a stochastic parrot.

Again, I think we give our own level of free will far too much credit. IMO we're not that far off from stochastic parrots. I mean most of you are on the higher side of the IQ spectrum, you haven't noticed at all that most people are archetypes? Our complexity makes us over value our own special snowflake version of reality. Yet try changing a pattern you've had for decades and watch the chemical breaks grind your will to a halt.

I was aware of more than 'most people are archetypes' as a teen, I saw most people as copies upon copies, not quite parrots but you won't see any metacognition there, and interrogating why they think this or that, most of the time it's going to simply boil down to what they've been trained to "think". I would not say that I'm different owing to some higher powered brain, I have to say it's about will.

As to that last sentence it's too reductive by half (I think you meant chemical brakes). "changing a pattern" as applied to any "you" (as per a universal you) that isn't yourself looks... overconfident to me. "grind your will to a hault". Nope. You can leave me out of that.

In any case you absolutely have not addressed "the problem of will in sentience" with that gesture. Again, I don't know what is even interesting about '[there is no] free will' and repeating this term is not solving any problem at all here. You seem confident that this 'we're not far off from the stochastic parrot' is supported by 'giving our free will too much sway' but all I have from that is wanting "will" to be obviated by the red herring "free will".

Repeated iterations and now this particular phrasing makes me wonder if you're not resisting the very idea of will. A pattern starts to appear out of a desire to believe something about the Large Language Model that isn't in evidence.

maybe AI never gets to what we would call independent thought, but it's already going to be by the nature of [what] it gives back to people, changing in some negative ways the way that humans behave.

It already has, albeit I wouldn't say it gives back anything, it regurgitates what there is.
One might want to entertain the notion "AI" might develop "Independent thought" (the qualifier independent is redundant, if it isn't independent it's just a rehash at best) but right now all there is is the Large Language Model (and artificial "neural network"), which doesn't engage with thought in any meaningful way. Thinking means reasoning. It isn't able. The current models are not going to evolve into a thinking engine, they're just not built for anything of the sort.

[BONES]

Again, I think we give our own level of free will far too much credit. IMO we're not that far off from stochastic parrots. I mean most of you are on the higher side of the IQ spectrum, you haven't noticed at all that most people are archetypes? Our complexity makes us over value our own special snowflake version of reality.

I think it's probably only us, most people are far too self-absorbed to notice much about anyone else.

Rick Rubin on the 60 Minutes TV program: "I have no technical abilities and I know nothing about music." Yeah here's someone we need listen to opinions from

Exactly, someone unconstrained by convention and dogma. Someone able to see things differently and create new paradigms no-one else had thought to try. Or maybe just someone who is slightly self-effacing, given his achievements. I mean, you don't get listed as one of the 100 Most Influential People in the World by Time magazine unless you know a thing or two about something. Mind you, Donald Trump regularly makes that list, so it's not infallible.

[jancivil]

Ok, so this word will has us at cross-purposes to begin with. I mentioned noticing people as though copies and nowt but, which was one of the stark realizations for me from taking LSD, which I started doing at age 14. People are taught or conditioned by all sorts of factors to be afraid of individuating, you fit in, you go along to get along.
As pertains to this word will, this fear of and failure if not refusal to individuate from first of all your parents means the will is weak or has been stifled. My mind was blown, I was out there where all of the bullshit disintegrates. There may be a path of return, for looking to retract and retreat for someone else but I was gone.

Another way to put this idea of will is the idea of being conscious of the moment in the moment, not second-guessing or where your ego and defense mechanisms raining all around you keep you from it. You are here now; you are a product of many things establishing the current version of you, your mind is a function of your brain, you brain is part of your body but you are not these things. You're probably acutely aware of your body but you are not your body. You are the experience of it; there is one thing we can be sure of in existence and that is consciousness. You may be dreaming, you may be crazy, you may have non-trivial questions about the nature of it all, but all of it is consciousness manifesting itself. You can't talk about the waves of an ocean as objects in the ocean, so no one is breaking consciousness down to component parts.

So what is consciousness? The sort of moderator of a discussion at IAI put this question to Roger Penrose: "Do you think that consciousness is more fundamental than the physics (eg., the physical brain)?" Sir Penrose: "I don't think that. Actually I don't know what I think." But now he gets into it, the brain as far as physics understands it is computational. As far as he is concerned, consciousness is beyond that, the neural processes, the whatstits and wherefores can't explain it. And funnily enough they do touch on the Large Language Model, strictly from computational.

I don't think there's a way around this. There's no 'there', there. The problem isn't even really understood. There is no imparting consciousness or individual will to a computer program.
It cannot experience: we all experience the qualia of things: the color red is understood, we know what it's like. "AI" has no idea whatsoever, & you can't make it understand.

As to my taking exception to 'we're not far from being that stochastic parrot', I only bother with the massive amount of exertion and all the years of preparation to get to where I compose music because I am interested in this music qua itself. I don't function in a vacuum so there's going to be others that get off on it but I can't predict it or manage it, or aim to please particularly, it would be a waste of energy. You can describe a box, you can believe in the box but you can't fit me into it. It's a figment. There is such a thing as an original thought. A lot of the world works against it but it doesn't vanish due to its inconvenience.

I can make all sorts of things that refer to extant models, tradition, convention, styles but the real deal is when the idea is not modeled. The things that happen are products of the idea; they never happened before. The adjustments made work within their own rules.

in reality you don't know what you're going to think of until you think of it, you don't plan to have that thought. Behaviors are intextricable from conditions, thought can be absolutely free.

Also in something putting the brakes on a person changing as if we're by definition stuck in patterns like we're unconscious, I have to point out that some of us are not the same person we were at numerous points in development. It seems clear enough that this type that is stuck with the patterns imposed on them can be exactly the person they were at 16 at age 60 only somewhat more degraded by time. Will, versus not.

I have no idea of some chemical process that slams the brakes on that, because as far as I know it doesn't happen. AFAIK this is more speculative fiction.