The Sentient AI Trap

[Cookiegod]

Care to cite just one?

Bwahahaha. The guy who still adamantly refuses to elaborate on his definition of intelligence, refuses to explain the contradiction inherent to the tidbits of his definition given, the guy who not once presented anything of substance here and oscillates between "the brain is very simple" and "we don't know how the brain works", the guy who writes the following about AI:

We are also capable of creating AI that can find and implement new rules into itself; finding patterns in the chaos and associating those patterns with responses. At the time we limit those AIs to simple functions. However, if you let an AI, as babies are, be able to alter its existing rules you create endless possibilities. In time, it can learn about its own existence and start creating new rules revolving around that which would amount to sentience. Learning self preservation it would create stricter rules to defend itself.

So, AI is not just training to predict outcome y based on input x. There are already AIs that predict outcomes that have not been defined before. They look at a chaotic data and create their own parameters. y is not defined. Yet, y1, y2, y3 and so on are created based on the observations of the AI. Its right out of the Person of Interest but DARPA does have a real project on it, called KAIROS. The idea is to look at all the data available and find connections that we can not see. Now, you may say that in simple terms the program works to find similar items. Sure, you can hardcode it like that but it is also possible to, as KAIROS aims to to a degree, apply machine learning and let the AI figure out its own parameters and rules for what is similar and what is not.

No, that's not how AI works. And no, you completely bungled the one and only concrete example you've given in this thread (KAIROS).

So how about no? I have several papers even on my hard drive on that very topic as we speak. The randomness in the brain is an established fact, and I can supply more. Not that it's needed, since it's the established scientific consensus and there's plenty on it you can google yourself. So I won't do until you finally do your part, which you so far have refused to do even on the most basic level.

[PointOfViewGun]

You might be shocked to hear that we do not have access to the papers on your hard drive.

[Muizer]

Despite the unnecessary jabs, it's still an interesting discussion.

I'll try a different angle. Feel free to respond.

Let's assume the essence of both artificial and organic intelligences is that they are composed of input-output systems, functions basically.

One difference I see with current AI is that the outcome of 'designed' functions is fully in line with an intent/purpose behind the design. That isn't to say the outcomes would be predictable in practice (a trained AI will find and use patterns no human designer would have come up with) or theory (functions might include an element of randomness by design).

On the other hand, evolution has no interest in 'purifying' the input-output system to its bare essentials. Evolved brain functions are going to be open to and impacted by inputs that do not serve the primary role (what we might think of as the 'purpose', if there had been a designer). Drug induced hallucinations, for example. Or functions may have ended up serving multiple purposes, in which case optimization would have concerned a compromise.

If that's the case, then the notion that "training" of individual components, fundamental in present day AI, could ever lead to consciousness in a whole might be called into question. Consciousness could happen by accident, but not by design, unless we can actually present "consciousness" itself as training data.

Ok, just rambling I suppose. There's probably countless of philosophers and scientists who have written papers on it that are too difficult for me to even read. What do you expect though. It's a gaming forum

[Cookiegod]

You might be shocked to hear that we do not have access to the papers on your hard drive.

And you'll remain well outside any position to ask me to provide anything until the very unlikely event that you're going to do any of the very basic things requested from you dozens of times.

I'll try a different angle. Feel free to respond.

Let's assume the essence of both artificial and organic intelligences is that they are composed of input-output systems, functions basically.

One difference I see with current AI is that the outcome of 'designed' functions is fully in line with an intent/purpose behind the design. That isn't to say the outcomes would be predictable in practice (a trained AI will find and use patterns no human designer would have come up with) or theory (functions might include an element of randomness by design).

On the other hand, evolution has no interest in 'purifying' the input-output system to its bare essentials. Evolved brain functions are going to be open to and impacted by inputs that do not serve the primary role (what we might think of as the 'purpose', if there had been a designer). Drug induced hallucinations, for example. Or functions may have ended up serving multiple purposes, in which case optimization would have concerned a compromise.

If that's the case, then the notion that "training" of individual components, fundamental in present day AI, could ever lead to consciousness in a whole might be called into question. Consciousness could happen by accident, but not by design, unless we can actually present "consciousness" itself as training data.

This is where you and I get much closer to consensus. It will always come down to how one defines sentience. It is my belief that the things I associate with sentience, be it in humans or dogs, is very much something that is heavily influenced by the evolutionary history we went through, as well as body chemistry, as well as the fact that we humans remain constantly in flux in a way that a computer cannot, nor has any reason to emulate. For any task I can think of, there's pretty much always a far more straightforward path that any algorithm is almost certainly to stay far away from consciousness.

When I walk the dog, or visit the friends that already have children, it's always very interesting for instance to notice those sudden mood swings which we adults have learned to control much better. Like a kid getting angry, and then after a couple of minutes then completely shifting 180° and being happy again. The moment you look at it from the cellular chemistry angle it becomes something impossible to see. Something triggers the release of e.g. neurotransmitters, kid/dog reacts accordingly, cannot control itself. After a couple of minutes the levels have fallen under a threshold. Kid/dog switches mood instantly again.

The other thing being the random noise which is a very strong element of our brain. It's not a bug, it's a feature, and makes us transmit signals much faster. The brain is a masterpiece of signal processing. But the random noise is there, is a strong element, and our brains have accorded significant flexibility in its function because of it, with a staggering amount of inside and outside factors affecting the result. Our brains do not operate on the assumption that the statistical impossibility of things firing the same way happening twice. Never mind that we do not have "trillions of universe ages" to wait for that to happen. Apart from our lifespan being far shorter than that, our brains are undergoing constant change throughout it.

A computer will always be very different for that simple reason alone.

Never mind that it will never evolve any of the traits we evolved as a social species. There's no reason for an AI to evolve empathy, there's no reason for it to evolve morality, nor will it ever confront mental health issues such as anxiety or existential dread.

As much as Ex Machina and Blade Runner are master pieces which I wholeheartedly enjoy, AI can be trained to emulate human traits, such as LaMDA allegedly pontificating over morality, but it will never actually have them for the very simple reasons that it doesn't need to.

That, and again, the fact that the huge differences in structure and purpose making huge differences in function inevitable.

But all that will never do away with another very human trait, which is to anthropomorphise everything we can. As we have well seen in this thread, but also throughout the history of humanity, we have a strong desire to imagine the human like. But anyone who's able to suppress that desire for a moment and look at the task in a rational manner should maybe look at the movies, literature and art (since we don't have a workable definition of sentience in this thread here) with artificial sentience that inspires us so in a critical light. Look at what specific traits and behaviours we attribute to those robots that make us think: "Yep, that's sentient!" And then think for a second why the algorithm should have developed that function to begin with. For me that's one simple way to burst that bubble.

Ok, just rambling I suppose. There's probably countless of philosophers and scientists who have written papers on it that are too difficult for me to even read. What do you expect though. It's a gaming forum

To be fair, almost no one in this forum is still playing Total War

[PointOfViewGun]

And you'll remain well outside any position to ask me to provide anything until the very unlikely event that you're going to do any of the very basic things requested from you dozens of times.

You tried to make the conversation as toxic as possible; largely ignored my points, misrepresented what I said, refused to provide how you viewed the relevant definitions, and repeatedly projected the shortcomings of your assertions and argumentation onto me. Now you speak of what position I'm in. Cute.

When I walk the dog, or visit the friends that already have children, it's always very interesting for instance to notice those sudden mood swings which we adults have learned to control much better. Like a kid getting angry, and then after a couple of minutes then completely shifting 180° and being happy again. The moment you look at it from the cellular chemistry angle it becomes something impossible to see. Something triggers the release of e.g. neurotransmitters, kid/dog reacts accordingly, cannot control itself. After a couple of minutes the levels have fallen under a threshold. Kid/dog switches mood instantly again.

The other thing being the random noise which is a very strong element of our brain. It's not a bug, it's a feature, and makes us transmit signals much faster. The brain is a masterpiece of signal processing. But the random noise is there, is a strong element, and our brains have accorded significant flexibility in its function because of it, with a staggering amount of inside and outside factors affecting the result. Our brains do not operate on the assumption that the statistical impossibility of things firing the same way happening twice. Never mind that we do not have "trillions of universe ages" to wait for that to happen. Apart from our lifespan being far shorter than that, our brains are undergoing constant change throughout it.

A computer will always be very different for that simple reason alone.

Our brains do operate on the assumption that particular processes will generate the same intended results. Not sure what you're trying to argue there.

[Sir Adrian]

Actually they don't, and this not something new, a simple google search will bring results from last year and from many years ago. The electric impulses in the neurons can sometimes misfire or trigger other neurons. Moreover the result is altered by an ever-changing mix of chemicals.

If the brain was deterministic learning would be impossible and there would be no such things as randomly remembering something, lapsus, deja-vu, phantom sounds, phantom limbs, etc.

[PointOfViewGun]

Actually they don't, and this not something new, a simple google search will bring results from last year and from many years ago. The electric impulses in the neurons can sometimes misfire or trigger other neurons. Moreover the result is altered by an ever-changing mix of chemicals.

If the brain was deterministic learning would be impossible and there would be no such things as randomly remembering something, lapsus, deja-vu, phantom sounds, phantom limbs, etc.

None of those assertions make sense. Misfires are, well, misfires. They do not break the rule. Changing chemical make up merely alters the process that is used. I'm not sure how determinism affect learning. In fact, can you actually go beyond making strict assertions and provide actual arguments why they are as so.

[Sir Adrian]

Ok. Here's what we are going to do. Prove to us without a shadow of a doubt that the brain is deterministic. So far you've only made random unsourced claims. Prove your point.

[PointOfViewGun]

Ok. Here's what we are going to do. Prove to us without a shadow of a doubt that the brain is deterministic. So far you've only made random unsourced claims. Prove your point.

What we actually should do is to pay attention to what people argue. So far I've only challenged the claims about the brain being deterministic. Failing to substantiate your claims can not be covered up by diving into a straw man argument.

[Sir Adrian]

You claimed the brain is deterministic and have question everyone who said it wasn't, including me 2 posts above. Prove your claims or take the loss.

[PointOfViewGun]

You claimed the brain is deterministic and have question everyone who said it wasn't, including me 2 posts above. Prove your claims or take the loss.

I'm gonna ask you to quote me making that claim and you're gonna follow right away with a quote having me making that claim? Right? I can quote myself twice where I argued something different.

[Sir Adrian]

I don't see any substantiation here. Can you prove any of your claims? Like say for instance this one where you question the statement that the human brain is non-deterministc

How is the human brain non-deterministic? It's really hard to make that suggestion one way or the other when we don't exactly understand how the brain works. Just because the brain considers a large number of variables it doesn't make it non-deterministic.

If the brain is not non-deterministic, it's deterministic, yes? Or are you arguing there's a 3rd possibility?

Prove your claims and stop playing dumb.

[Septentrionalis]

If the brain was deterministic learning would be impossible and there would be no such things as randomly remembering something, lapsus, deja-vu, phantom sounds, phantom limbs, etc.

None of those things prove that the brain is non-deterministic. Are you suggesting it works randomly? Just because the brain is extremely complex in its finite states of processing does not prove that it is non-deterministic.

[PointOfViewGun]

I don't see any substantiation here. Can you prove any of your claims? Like say for instance this one where you question the statement that the human brain is non-deterministc
If the brain is not non-deterministic, it's deterministic, yes? Or are you arguing there's a 3rd possibility?
Prove your claims and stop playing dumb.

Clearly you lost track of what you were objecting and decided to make things up along the way. Questioning someone's claims doesn't really mean one argues one way or the other. I could be the number one proponent of the idea that brain is deterministic and still question it based on the arguments one makes. I don't need to take anything you say at face value. I'm not the one making a claim here. You are. It's on you to stand by it and support it. All of that, by the way, is just you trying to deflect from substantiating your assertions in the simplest manor. It's not a good look.

[Sir Adrian]

Substantiate your claims with sources Seth. If you don't want to substantiate this then how about your claim that humans are just complex calculators. Source it.

[Muizer]

Actually they don't, and this not something new, a simple google search will bring results from last year and from many years ago. The electric impulses in the neurons can sometimes misfire or trigger other neurons. Moreover the result is altered by an ever-changing mix of chemicals.

If the brain was deterministic learning would be impossible and there would be no such things as randomly remembering something, lapsus, deja-vu, phantom sounds, phantom limbs, etc.

I feel this is muddying the waters. At least, it is my understanding that determinism concerns the fundamental nature of reality, not our conceptualization of it and certainly not at the level of biological units.

[Sir Adrian]

I can see where you're coming from but when I talk about determinism vis-a-vis the brain I generally mean does the brain produce the same output for the exact same input.

This Forbes article should clear up what I'm arguing (and why Seth is wrong)

[Cookiegod]

largely ignored my points

I went through most of them point by point rofl, you failed to address pretty much all the counter points. After claiming that abaci weren't intelligent, you switched to saying that mechanised were. My question how the propulsion can make a difference when it works according to the exact same rules as a normal Japanese Abacus: Silence. Even though I brought it up multiple times to force an answer, which you refused to give.

misrepresented what I said

No, I tried to make your claims falsifiable, you refused to clarify, collaborate or elaborate. The "argument" that "an abacus is not intelligent because it isn't" isn't going to cut it.

refused to provide how you viewed the relevant definitions

Actually, I'm the one who provided multiple definitions from various sources to prove your claim that you were simply representing the common definition wrong. Actually, calling it a claim is too generous. This here was an outright lie:

The definitions I use are the ones existing in the dictionaries. I have been asking yours to see how it differs from those.

Still at this moment we're still chasing a ghost for the simple reason that you are refusing to define what the terms meant to you. I meanwhile have argued the entire range of sentience. From the reductionist take (TL;DR: It's meaningless) to what I would consider sentient (unfeasible and also pointless to emulate). But it'd be far better that a believer provide us with a good point to argue from, not one who denies it. Muizer has since done what you have failed at, and Muizers position has not suffered a bit from it while still moving the discussion forward.
Meanwhile you refused to provide us yours even after multiple pages of begging you for one:

Your continued refusal of making a clear stand, which quite frankly is the most likely outcome, just like it has been until now, on the other hand will be continued proof of a lack of clarity and argument.

If you do not know what sentience and intelligence is, the inevitable conclusion is that you are completely disqualified to judge whether artificial sentience is achievable or not.

and repeatedly projected the shortcomings of your assertions and argumentation onto me.

No, I presented you with a number of issues with your own baseline assertions. I did take it easy on you because your claim re KAIROS for instance is completely wrong, but it would be pointless to argue that if you don't even have a basic understanding of the topic at hand. This here was such a dumpsterfire that I stayed well clear of it:

A calculator knows what to do given the parameters and functions installed in it. In the most strict sense it has an artificial intelligence. It's just at a very basic level. Then you have a car that can drive itself. It has a numerous number of inputs coming from multiple sensors in real time that checks a high number of parameters executing a high number of functions. You could consider them having a medium level AI that doesn't have the capability to learn (though there are some projects that aim just that). We are also capable of creating AI that can find and implement new rules into itself; finding patterns in the chaos and associating those patterns with responses. At the time we limit those AIs to simple functions. However, if you let an AI, as babies are, be able to alter its existing rules you create endless possibilities. In time, it can learn about its own existence and start creating new rules revolving around that which would amount to sentience. Learning self preservation it would create stricter rules to defend itself.

So, AI is not just training to predict outcome y based on input x. There are already AIs that predict outcomes that have not been defined before. They look at a chaotic data and create their own parameters. y is not defined. Yet, y1, y2, y3 and so on are created based on the observations of the AI. Its right out of the Person of Interest but DARPA does have a real project on it, called KAIROS. The idea is to look at all the data available and find connections that we can not see. Now, you may say that in simple terms the program works to find similar items. Sure, you can hardcode it like that but it is also possible to, as KAIROS aims to to a degree, apply machine learning and let the AI figure out its own parameters and rules for what is similar and what is not.

Now you speak of what position I'm in.

Yeah, it's not a good one.

Our brains do operate on the assumption that particular processes will generate the same intended results. Not sure what you're trying to argue there.

I'd normally ask you to back up your claim that your brain always will generate the same intended results. But we both know that you will not do such a thing. I will thus simply contend that with a simple task, such as 1+2=3, your brain will with a very high likelihood arrive at the same conclusion. For more complex, out of routine and spontaneous decisions, I do not share your view.

So far I've only challenged the claims about the brain being deterministic. Failing to substantiate your claims can not be covered up by diving into a straw man argument.

The challenge: "We don't actually know how the brain works". After claiming for multiple pages that the brain is in essence a slightly more complicated calculator:

You, as a human being, is not different from a calculator, in principle. A calculator has its inputs, rules and output mechanism. Same as you. You merely differ in complexity.

How is the human brain non-deterministic? It's really hard to make that suggestion one way or the other when we don't exactly understand how the brain works. Just because the brain considers a large number of variables it doesn't make it non-deterministic.

And how do you resolve the glaring contradiction between those two takes? Never mind, I forgot for a second we're not allowed to ask you that

None of those assertions make sense. Misfires are, well, misfires. They do not break the rule.

The way you're talking about misfires here shows very clearly to me that you still haven't grasped the basic fact that the "misfires" aren't a bug, but a feature. It speeds up the processing in your brain. Any assertion to the contrary I'd request you to back up. But unfortunately you never will.

None of those things prove that the brain is non-deterministic. Are you suggesting it works randomly?

The brain does have a very high degree of randomness baked into it, yes. If one wants to then argue the causality of the entire universe that's up to them, but the point is very simple: Your brain as a selfcontained unit is influenced by a huge variety of inside and outside factors affecting the processing and outcome.

I feel this is muddying the waters. At least, it is my understanding that determinism concerns the fundamental nature of reality, not our conceptualization of it and certainly not at the level of biological units.

Not to doxx myself more than necessary, but while I am not a neuroscientist, I have written both a thesis and a conference paper in the field of signal processing, which our brains do a lot. And no one so far has objected to the word "Random" appearing even in the title, or starting to argue the causality of the universe as a whole.

There is a simple reason why I wrote non-determinism instead of free will, which I have already mentioned before: Free will is not something that matters to our discussion here, and is a completely different can of worms. It is not even a topic that interests me particularly.
But unless one wants to claim that the brain factors in the state of every single particle in the entire universe, or is predicated on functioning only in the statistically impossible event that the brain arrives at the exact same position as before, then the brain as a self contained unit has to work within the context that there's a huge variability in its processing, which makes it non-deterministic and non-reproducible.

If any of you want to keep harping on about quantum theory and the causality of the universe, even though it's not relevant to the discussion here, then please embrace the words of these two great minds:

God doesn't dice.

Don't tell god what to do, Einstein.

[Muizer]

There is a simple reason why I wrote non-determinism instead of free will, which I have already mentioned before: Free will is not something that matters to our discussion here, and is a completely different can of worms. It is not even a topic that interests me particularly.
But unless one wants to claim that the brain factors in the state of every single particle in the entire universe, or is predicated on functioning only in the statistically impossible event that the brain arrives at the exact same position as before, then the brain as a self contained unit has to work within the context that there's a huge variability in its processing, which makes it non-deterministic and non-reproducible

You're saying non-determinism is due to what we choose to not factor in our predictions, rather than what cannot be factored into predictions? That would make the assertion of non-determinism quite arbitrary, no?

[Cookiegod]

You're saying non-determinism is due to what we choose to not factor in our predictions, rather than what cannot be factored into our predictions?

I don't understand your question. Rephrase it please.