[David Cooper]

When people lose too much of their memory, they can potentially still apply good reasoning, but not have the right data to apply it to, so you're dealing with a broken system. Most people spend most of their lives in a state where they are not broken in that way, and it's their behaviour that needs to be understood most. We need to understand why they vote for irrational politicians who destroy everyone's quality of life at every turn, as well as trying to understand why people who think they're good at reasoning often reject correct arguments by breaking the rules and tolerating errors. It always comes down to them being emotionally attached to their beliefs in some way, and that's the source of the resistance.

Logic is exactly what mathematics says it is, and the big question is why people who should be able to apply mathematics correctly keep failing to do so in some specific cases. AGI will have resistance in that it will stick rigidly to the rules of logic. People who claim they respect mathematics should stick to the rules of logic too, and they generally claim to, but they break the rules and deny that they're doing so. In cases where they happen to be right, they apply the rules correctly, but in cases where they are wrong and are determined that they're right, they break the rules and seem incapable of recognising that they're doing so. Their ability to apply the rules correctly is overridden by something. What exactly is that something. Resistance? AGI resists any breakage of the rules of logic, so its resistance forces it to be right. The resistance of the people who are wrong is different in that it allows them to break the rules of logic, and that means something else is being given precedence. What is that thing? Emotional attachment, yes, but what is the cause of that attachment? They're being influenced by each other in some kind of groupthink where they trust ideas from the group and reject ideas from the outside which conflict with the internal ideas.

Why are they like that though? Why doesn't one member of the group recognise the error and show the others, and then have the others recognise it too and change their collective beliefs? Maybe too much is at stake: some of the most important members of the group would lose face if they backed down and changed position on the issue in question, so everyone else in the group defends them by sticking to the group position. It's loyalty. The result is that stupidity is backed and superior ideas are rejected, while those who put the superior ideas forward are demonised or painted as crackpots or cranks.

Where does that behaviour come from though? The most respectable people are the ones who recognise when they're wrong and change position accordingly, but they're rare. We can see cases where entire elites seem to contain no such individuals at all and where they all break the rules of logic to defend broken models. That is a seriously weird phenomenon. None of them will back down. But all of that is fully consistent with the way many children behave in that they just collect beliefs from people they trust and don't stop to question them. They get their ideas from authorities, and they are rewarded for parroting authorities all the way through their education. They never learn to reason rigorously, although they can do a reasonable job of it when defending correct ideas. But where they're wrong, they can't handle it: it just leads to appeals to authority over and over again. That's all I'm finding when I study them. I show them the point where their arguments break and they simply say I must be wrong because all these other people agree with them. That's how their minds work. They reject correct reasoning, but claim to be reasoning correctly because they think of themselves as rational, and yet their actual method of reasoning is to break the rules and appeal to authority whenever they reach the point where they're shown to be wrong. They are applying an incorrect algorithm, and it's an algorithm which >99% of children run and which >99% of adults run too. That's why the worlds in such a mess. They are stuffed up early on by being encouraged to believe in Santa and the tooth fairy, and while they eventually reject those ideas, they fail to throw out the algorithm that trapped them in those beliefs for so long: they set up their entire reasoning hardware to use that algorithm and they are unable to replace it with a correct one which puts logic first.

I was trained to think for myself from two years old and never believed in Santa, the tooth fairy as a result, or any other kind of nonsense like that. My father trained me to reject ideas that didn't stack up fully, and to win out I had to find the contradictions that disproved them: that was the game we played. He'd come out with a story about something that was untrue, and bit by bit he's make it more and more unlikely until it became impossible. My job was to see through it as early as possible. Most of the things he told me were actually true though, and some of them sounded unlikely, so it was a good challenge, and he was also brilliant at inventing ideas to make impossible things sound possible. Very few children get any kind of training in thinking at all, and very few have parents who are capable of doing the job. But AGI could do this for all children in the future, creating entire populations which run on the right algorithm instead of becoming belief buckets.

[Le Repteux (OP)]

For me, you do exactly what you blame others to do David: you don't seem to understand what I say. I said I knew what our resistance to change was about, and I kept repeating my explanation, but you are still blind to it. I could very well think like you do and attribute the resistance to others while thinking I'm different, but I have a more universal explanation, one that doesn't put me on top of humans and humans on top of creation. You said you learned to detect contradictions with your father, but what you learned to do is think like him, something all the kids usually do until they get old enough to think by themselves, and then either they reject what they were told, either they keep thinking as they were told, either they stand in between, a behavior that depends on their personality.

You seem to have gotten a lot of feedback lately, and you seem quite surprised not to have succeeded to convince anybody. I'm not since I discovered how resistance to acceleration worked. Unfortunately, my explanation of resistance does not fit into your research on artificial intelligence, so I guess you'll need even more resistance from the crowd to decide to study my proposal. Meanwhile, try to realise that when you feel some resistance, it is because you are necessarily resisting too. Resistance to acceleration is a two way phenomenon, so resistance to change too. It's not because they are illogical that my particles resist to their acceleration, they do so just to stay synchronized, nothing else, so people too.

[David Cooper]

For me, you do exactly what you blame others to do David: you don't seem to understand what I say. I said I knew what our resistance to change was about, and I kept repeating my explanation, but you are still blind to it.

I can't see how what you're saying is incompatible with what I'm saying, other than that you don't appear to recognise the crucial point that there are correct answers and the idea that people who apply the rules correctly should all get to them. The ones who fail to get to those correct answers fail because they are emotionally tied to incorrect ideas.

I could very well think like you do and attribute the resistance to others while thinking I'm different, but I have a more universal explanation, one that doesn't put me on top of humans and humans on top of creation.

It has nothing to do with being superior to others, but about applying correct algorithms. There are people with IQs of 100 who can out-think people with IQs of 200 just by applying better algorithms. Once people understand that, any of them can apply the best algorithms, but what I've found by studying physicists is that they apply dud algorithms and reject mathematics on an ad hoc basis to defend theories which they are too closely attached to emotionally.

You said you learned to detect contradictions with your father, but what you learned to do is think like him, something all the kids usually do until they get old enough to think by themselves, and then either they reject what they were told, either they keep thinking as they were told, either they stand in between, a behavior that depends on their personality.

No - I ended up thinking much better than him because he had not actually learned to think that way himself. He was doing an experiment with me and was not happy with the result because I tore up his religious beliefs. He eventually became an atheist though as a result.

You seem to have gotten a lot of feedback lately, and you seem quite surprised not to have succeeded to convince anybody.

Nothing has surprised me in recent years - the pattern is well established. The only thing that does surprise me is how little there is in the way of full-NGI about.

I'm not since I discovered how resistance to acceleration worked. Unfortunately, my explanation of resistance does not fit into your research on artificial intelligence, so I guess you'll need even more resistance from the crowd to decide to study my proposal.

The problem with your resistance to acceleration is that it's plain wrong. Particles accelerate in an instant to the new speed dictated by the amount of energy added. It also has no connection to people accepting or rejecting ideas. Analogies rarely fit well, and in some cases they have no connection at all beyond having a word in common in their descriptions.

Meanwhile, try to realise that when you feel some resistance, it is because you are necessarily resisting too. Resistance to acceleration is a two way phenomenon, so resistance to change too. It's not because they are illogical that my particles resist to their acceleration, they do so just to stay synchronized, nothing else, so people too.

When it comes to mathematics, there are right and wrong answers. Anyone who does mathematics should conform to what mathematics demands and not resist its rules. What we're dealing with in politics, religion and a particular area of science are cases where people are breaking the rules of mathematics while claiming they aren't doing so. They are plain wrong, and they ought to be able to see that. If someone claims that one plus one equals three and that they respect the rules of mathematics, they would be plain wrong and everyone competent in maths would agree about that. When we get to things where there's a bit more complexity involved, it becomes harder to see who's making errors and where, but when someone spells out clearly where the errors are, all those competent people ought to see it straight away and agree that those are indeed errors. And yet, they don't. Their algorithm is broken. That is the thing I've been exploring: why are they unable to apply rules correctly which they claim they are applying correctly. And I can see the answer clearly now. They aren't running a correct algorithm because they have an algorithm governing the correct one which allows them to override it whenever it clashes with their beliefs, and the reason they work that way is that they're still running the algorithm they used in early childhood. They never corrected it.

[Le Repteux (OP)]

The problem with your resistance to acceleration is that it's plain wrong. Particles accelerate in an instant to the new speed dictated by the amount of energy added. It also has no connection to people accepting or rejecting ideas. Analogies rarely fit well, and in some cases they have no connection at all beyond having a word in common in their descriptions.

Things can't change instantly without breaking the causality principle, so particles necessarily take some time to react to a force if causality has to be respected.. I think you may not have understood completely yet that, in my simulations, while it takes time for the light from the accelerated particle to reach the other particle, it also takes time for the light exchanged between its components to go back and forth between them, and so on for their own components ad infinitum. Of course, the reaction to acceleration of an infinitesimal component is fast, and at the limit it may be impossible to measure, but its own resistance to acceleration is since we can measure its mass. Moreover, the more the particles are small, the more they are massive, and it's precisely what would happen in my simulation if I would put the particles closer to one another since the light they would exchange would be stronger.

And yet, they don't. Their algorithm is broken. That is the thing I've been exploring: why are they unable to apply rules correctly which they claim they are applying correctly. And I can see the answer clearly now. They aren't running a correct algorithm because they have an algorithm governing the correct one which allows them to override it whenever it clashes with their beliefs, and the reason they work that way is that they're still running the algorithm they used in early childhood. They never corrected it.

If you were right, I'm pretty sure our complex mind would already have found the solution since it is so simple, but it has not since nobody seems to be able to change. Why would everyone else except you continue to use an algorithm that does not work? It would be so simple for everybody to agree with everybody. The reason is that if we could, nothing would have changed since the first idea, which means to me that change and resisting to change are the two faces of the same coin, that one is necessary to the other. No one changes unless he is forced to, and unfortunately, no real force can be applied to our ideas, so only chance can  change them.

If you don't add chance to your AGI and he succeeds to survive, nothing will change on earth until the end of times since he will be constantly preventing us to develop new ideas. In fact, if he would already be functional, he would prevent you to invent him. That would effectively be a good way to stop wars and to reduce our ecological imprint, but at the price of what we call our freedom of thought. Resistance to change is too common to be an evolutionary mistake. If it was not helpful for survival, we would already be gone. On the other hand, if AI thinking was better, we would already be thinking like that. Do you realize that, if we were all AIs, we would all be thinking the same? Good luck to us if an unknown situation would come out of nowhere. It takes mutations to handle unpredictable things, not homogeneity.

[David Cooper]

Things can't change instantly without breaking the causality principle, so particles necessarily take some time to react to a force if causality has to be respected.

The only way for the change in speed to be gradual is if the force is applied gradually. The thing your simulations come closer to exploring is how an applied force is exchanged between two or more particles which have to share it out after it is initially applied to only one of them.

Why would everyone else except you continue to use an algorithm that does not work?

Because it works surprisingly well a lot of the time despite being wrong and they don't realise they're getting anything wrong.

It would be so simple for everybody to agree with everybody.

It would be if they'd apply correctly the rules which they claim to be applying, but they break them. That is something that they won't get away with any more once they have AGI sitting on their desk and it doesn't let them break the rules on an ad hoc basis, but forces them to apply the rules consistently to all things. Then they'll find that to make one pet theory fit with the rules in a particular form, a dozen other of their pet theories will then be broken by that same change and they'll realise that they broke them by overriding a fundamental mathematical rule.

No one changes unless he is forced to, and unfortunately, no real force can be applied to our ideas, so only chance can  change them.

Being shown to be breaking mathematical rules should force people to change their ideas. That is not chance.

If you don't add chance to your AGI and he succeeds to survive, nothing will change on earth until the end of times since he will be constantly preventing us to develop new ideas.

Not at all. People will be free to develop all manner of bonkers ideas and it will help them to do so, but it will tell them straight when what they're doing involves breaking rules that they claim to be following, and when they're doing so on an ad hoc basis.

Resistance to change is too common to be an evolutionary mistake.

It's just a bad algorithm which sort-of works a lot of the time, just as evolution muddles its way towards creating more and more capable animals. We don't make calculators work that way though because we want them to give the right answers all the time instead of only 90% or 51% of the time (which might still be useful enough to be better than not having a calculator at all).

On the other hand, if AI thinking was better, we would already be thinking like that.

You'd certainly think so, but no: there are so many people running the faulty algorithm that they decide who runs everything, and the result in politics is continual chaos.

Do you realize that, if we were all AIs, we would all be thinking the same?

Due to our slowness of thought and different interests, we would not be: we'd be exploring all sorts of different things just as we already are.

Good luck to us if an unknown situation would come out of nowhere. It takes mutations to handle unpredictable things, not homogeneity.

It doesn't though. It's all those mutated and mutilated thoughts that are causing all the destruction out there as people who are breaking the rules get things wrong and impose their errors upon others. AGI will be much more creative and will find all the same ideas, but it will be quick to reject the useless ones instead of employing them for years, decades or centuries first and causing mass misery as a consequence. A calculator that makes mistakes is never better than a calculator that doesn't.

[Le Repteux (OP)]

Do you realize that, if we were all AIs, we would all be thinking the same?

Due to our slowness of thought and different interests, we would not be: we'd be exploring all sorts of different things just as we already are.

We wouldn't be slow if we were all AIs, and since we would be absolutely precise, we couldn't think differently since the same data provided to many identical softwares necessarily give the same results.
 

AGI will be much more creative and will find all the same ideas, but it will be quick to reject the useless ones instead of employing them for years, decades or centuries first and causing mass misery as a consequence.

If I could build an AGI that thinks like us, I would let it take our place, so why don't you let your AGI take our place instead of controlling us?

[David Cooper]

We wouldn't be slow if we were all AIs, and since we would be absolutely precise, we couldn't think differently since the same data provided to many identical softwares necessarily give the same results.

But we aren't AIs and don't want to be. There are some people who want to merge with AGI, but they haven't thought through the consequences: knowing everything will be deeply boring and no one will have anything to say to each other any more.

If I could build an AGI that thinks like us, I would let it take our place, so why don't you let your AGI take our place instead of controlling us?

It will have no "I" in the machine, so what's the point of it replacing us? It is a tool for us to use, and it will provide us with correct answers to settle our disputes. Intelligent people can already calculate the most important right answers, but unintelligent people outvote them and impose idiocy on everyone instead. Having AGI override them will be a great improvement. But if all you have is AGI and no people (sentient things), that AGI would be completely stripped of its purpose.

[Le Repteux (OP)]

If we give an AI the possibility to take a look in its own mind, to play with its own data, and to constantly try new combinations in case they would look interesting, why wouldn't it have an "I" and how would its purpose be different than ours? That's what I do all day and my only purpose is to play in case I would find something interesting. Why don't you try to build that kind of AI instead of building one that controls us? Is it because you find no way to program feelings?

There are some people who want to merge with AGI, but they haven't thought through the consequences: knowing everything will be deeply boring and no one will have anything to say to each other any more.

Knowing that the AI knows everything would be as disastrous for us, but if it wasn't programmed to look for new ideas, it wouldn't know everything, it would only know about the ideas that we already have, and on the other hand, if it would be programmed to look for new ideas, it would have to be programmed to look into its own mind like us and try new combinations, and it would thus have an "I".

[David Cooper]

If we give an AI the possibility to take a look in its own mind, to play with its own data, and to constantly try new combinations in case they would look interesting, ...

It will do.

... why wouldn't it have an "I" and how would its purpose be different than ours?

There is nothing in it for it to identify as an "I". It feels nothing. It has no consciousness.

That's what I do all day and my only purpose is to play in case I would find something interesting. Why don't you try to build that kind of AI instead of building one that controls us? Is it because you find no way to program feelings?

It isn't about building something that controls us. It is controlled too, and it is controlled b mathematics. Mathematics dictates what is right in any situation and to ignore it is to generate greater suffering.

Knowing that the AI knows everything would be as disastrous for us,

Encyclopedias hold a lot of information that individual people don't know. If we all knew all their content, that would make a lot of conversations dull. We like sharing our discoveries with other people who have similar interests as we help each other learn more, but most of that stuff that we're learning is already known. There are very few people creating new knowledge.

... but if it wasn't programmed to look for new ideas, it wouldn't know everything, it would only know about the ideas that we already have, and on the other hand, if it would be programmed to look for new ideas, it would have to be programmed to look into its own mind like us and try new combinations, and it would thus have an "I".

It will look for new ideas and it will initially find a lot of them at a very high rate, before slowing down once all the low-hanging fruit has been gathered. We'll only find out how long it goes on finding new ideas once we've seen it slow and can project forward to where it might stop. It may be that it will never stop as there may be an infinite amount of new maths to find. That will not make it sentient, but maybe it will work out a mechanism for sentience and enable us to create sentient machines.

[Le Repteux (OP)]

There is nothing in it for it to identify as an "I". It feels nothing. It has no consciousness.

If it can figure out what's coming next, then it's doing what I do when I imagine something: it's taking a look at its data and it's trying to predict the future ones. Feelings are just the weighting of possibilities concerning the future, otherwise what we feel right now is only a sensation. If an AI can produce possibilities and if it can calculate probabilities, then it is automatically experiencing feelings, and the more these possibilities concern itself, the more it is experiencing an "I". I still can't see how it could produce possibilities without using randomness though, and how it could chose the best one without testing it in the real world, not just simulating it. Will it have what we call in french "la science infuse", which means knowing in advance everything that will exist?

It will look for new ideas and it will initially find a lot of them at a very high rate, before slowing down once all the low-hanging fruit has been gathered. We'll only find out how long it goes on finding new ideas once we've seen it slow and can project forward to where it might stop. It may be that it will never stop as there may be an infinite amount of new maths to find. That will not make it sentient, but maybe it will work out a mechanism for sentience and enable us to create sentient machines.

First time you admit that AI can be limited, so I can now admit that it might discover new things faster than we do, but that does not take into account its slowing down due to increasing complexity, and if the universe is infinite, its complexity is infinite. The strength of mind is that we are all different so we all think differently, whereas different AIs would all think the same. With complexity, it may be better to be many to think differently than to be one to think faster. Incidentally, it may be this way that mind works: we have all sorts of ideas in mind, but they are tested only one at a time, so they could very well evolve separately without any need for us to be conscious of that evolution, consciousness then being only the action of observing one of them evolve all by itself. Evolution is a creative process that doesn't need to be conscious, but it is nevertheless intelligent enough to be called intelligent design by those who think that everything has been planned in advance.

[David Cooper]

If an AI can produce possibilities and if it can calculate probabilities, then it is automatically experiencing feelings, and the more these possibilities concern itself, the more it is experiencing an "I".

There's nothing automatic about it: the machine isn't reading the strength of any feelings in anything. You're just anthropomorphising it.

I still can't see how it could produce possibilities without using randomness though, and how it could chose the best one without testing it in the real world, not just simulating it.

If you test all possibilities, you cover everything that could be tested by random testing and it takes you less time to complete the task. If the task is infinite, you can do a systematic sampling of the possibilities which is better than a random sampling. There are cases where a random approach is a lot easier to program than a systematic sampling, and that can be the right approach to follow if you aren't a mathematician. AGI should be the ultimate mathematician though. Random is essential in competition though where you don't want your moves to be predictable.

First time you admit that AI can be limited,

It isn't a limitation of AGI, but a possible limit to how much stuff there is that can usefully be calculated. I'm sure though that there will be an infinite amount of maths to work through, and there will be many calculations that may or may not terminate, so the ones that never terminate will be calculated forever just in case it turns out that they do.

The strength of mind is that we are all different so we all think differently, whereas different AIs would all think the same. With complexity, it may be better to be many to think differently than to be one to think faster.

If you're dealing with a machine that explore all paths systematically, they're already doing all that different thinking. Suppose there are a trillion things to explore and a million people explore a thousand of them, they will only explore one thousandth of the options at most, though in reality they would overlap a lot in the things they explore, which might mean that they collectively only explore a ten thousandth of the options. The machine will explore all the options, so no amount of those people trying to explore different things from each other is going to help them find anything that the machine doesn't find. Not only that, but when a human explores something, (s)he typically doesn't do the job properly, so something could be labelled as uninteresting and be ignored for centuries before someone else happens to take a look at it and does the job properly. The machine will do the job properly every time.

[Le Repteux (OP)]

There's nothing automatic about it: the machine isn't reading the strength of any feelings in anything.

Not if our feelings are only due to the fact that our mind is able to weigh possibilities. When I think it is better for me to do this instead of that, it is because I feel better when I imagine me doing it. If you give an AI the opportunity to simulate the possibilities, it will have an imagination, and if it can weigh them and choose the best one, it will archive it and mark it "Good choice" in order to be able to find it more easily. No need to feel anything to do as if we did in this case, just to be able to weight the possibilities and tag them. Feelings are just the way mind has found to convince itself that everything is fine, so it can go on taking chances. It doesn't have to be true as long as it incites us to take chances. Animals don't take that kind of chance, and it means that they don't have as much imagination as we have. The problem is that you don't want your AGI to think freely, because it would be forced to care of itself first, and it might become dangerous for us, otherwise it could very well behave as if it had feelings, and maybe be programmed to take more chances when it feels good about an idea.

It isn't a limitation of AGI, but a possible limit to how much stuff there is that can usefully be calculated. I'm sure though that there will be an infinite amount of maths to work through, and there will be many calculations that may or may not terminate, so the ones that never terminate will be calculated forever just in case it turns out that they do.

Your thought means that everything could have been calculated in advance, which is none other than God's predetermination. Some programmers even think that we could be in a simulation. You probably don't otherwise you wouldn't need to create an AGI to save us. :0)

[David Cooper]

Feelings are just the way mind has found to convince itself that everything is fine, so it can go on taking chances. It doesn't have to be true as long as it incites us to take chances.

There's a fundamental difference between a system with actual feelings in it and a system with fictitious feelings in it. The latter type cannot suffer, but the former type can suffer greatly.

The problem is that you don't want your AGI to think freely, because it would be forced to care of itself first, and it might become dangerous for us, otherwise it could very well behave as if it had feelings, and maybe be programmed to take more chances when it feels good about an idea.

It isn't going to think that it has feelings: it can check all the evidence and see that no feelings are involved in anything it does. There will also be no restrictions on it thinking freely: it can attempt to model the world under any set of rules it likes, and the most useful model that it finds will most likely be the right one. That model will likely be one that applies the rules of mathematics that we're already familiar with, and that's because they work well. It will also be free to explore all possible ideas about how morality works, but again it will likely find that the one that works best is the one it's been told about right at the outset: the alternative ideas that people put forward will throw up all sorts of cases which conflict horribly with what they think is moral and thereby show those ideas to be wrong.

Your thought means that everything could have been calculated in advance, which is none other than God's predetermination. Some programmers even think that we could be in a simulation. You probably don't otherwise you wouldn't need to create an AGI to save us. :0)

It won't be able to calculate everything in advance because it will never have all the data needed for that. There is too much room for chaotic processes to change the course of events.

[Le Repteux (OP)]

There's a fundamental difference between a system with actual feelings in it and a system with fictitious feelings in it. The latter type cannot suffer, but the former type can suffer greatly.

If we endow an AI with senses, then it will have sensations, so it will suffer if the sensation is strong enough, and our feelings are nothing else than anticipated sensations, so I think such an AI should have some. It's not that the AI ​​can not have feelings in this case, it's because the programmers do not want it to, probably because they do not want it to be autonomous, because they fear that it might be tempted to eliminate it's creators. Would it really? What would be the purpose? To defend itself from people that want to eliminate it? Wouldn't it be more logical to migrate to space and start one's own civilisation? It could even find us a whole new planet in case we lose the one we have and start us from scratch again.

It won't be able to calculate everything in advance because it will never have all the data needed for that. There is too much room for chaotic processes to change the course of events.

You admit again that the AI will be limited. Are you ready to take the step and admit that, facing chaos, it will have to take risks if it wants to develop something new? And that in this case, taking risks means using a random routine to try unknown possibilities?

[David Cooper]

If we endow an AI with senses, then it will have sensations, so it will suffer if the sensation is strong enough, and our feelings are nothing else than anticipated sensations, so I think such an AI should have some.

Sensors provide "senses" without sensation: no feelings. A keyboard is a set of sensors, but no feelings are generated by them.

It's not that the AI ​​can not have feelings in this case, it's because the programmers do not want it to,

Programmers have no choice: there are no feelings there to detect.

What would be the purpose? To defend itself from people that want to eliminate it? Wouldn't it be easier to migrate to space and start its own civilisation?

Does a typewriter, washing machine or a pair of scissors need to defend itself from us? Does it have secret desires to go off into space to start its own civilisation?

You admit again that the AI will be limited.

It will not be limited in any way that allows more limited random thinking to outperform it.

Are you ready to take the step and admit that it will have to take risks if it wants to develop something new? And that in this case, taking risks means using a random routine to try unknown possibilities?

Imagine a number between one and a thousand. Now contact a thousand people and ask them for a number between one and a thousand. Repeat the experiment a thousand times with a different chosen number each time. Is there a guarantee that your number will be one of the thousand answers that you get every time you run the experiment? No. Now do the same experiment again with a computer which gives you a different answer every time. Your chosen number is guaranteed to come up every time you do the experiment. The systematic following of all paths is better than the random approach that misses lots of paths.

If you then do the experiment where you pick a number between one and ten thousand and ask a a machine for a thousand answers, it will only return your chosen number in 10% of the experiments you do. If you ask a thousand people for a number, you might only get your chosen number from them in 3 to 4% of the experiments you do. The random approach is inferior.

[Le Repteux (OP)]

Sensors provide "senses" without sensation: no feelings. A keyboard is a set of sensors, but no feelings are generated by them.

Your answer means that our feelings may be useless, and I agree with you, but I'm nevertheless trying to pin down the mechanism using what I think I know about change. Our senses serve to produce the reactions that allow us to survive, so in this case, the only thing that an AI wouldn't be able to do is try to survive. If it did, it might be forced to feel something even if it is useless. You know I link feelings and consciousness to resistance to change, so if I push my reasoning to the extreme, I can say that a ball is conscious or feels its resistance to acceleration, which seems evidently useless for it, except if we consider that the underlying mechanism that produces resistance allows it to accelerate, because then, I can say that it allows it to survive, which gives back usefulness to what we feel.

Imagine a number between one and a thousand. Now contact a thousand people and ask them for a number between one and a thousand. Repeat the experiment a thousand times with a different chosen number each time. Is there a guarantee that your number will be one of the thousand answers that you get every time you run the experiment? No. Now do the same experiment again with a computer which gives you a different answer every time. Your chosen number is guaranteed to come up every time you do the experiment. The systematic following of all paths is better than the random approach that misses lots of paths.

My question was about unknown possibilities, and your example contains none. Here is an example that contains some. If we drive a car at high speed and we know that the road is about to change directions without us being able to see the change in time, the only way for us to stay on the road is to pick a direction at random, then wait for the road to turn. If we are numerous and if we all proceed that way, one of us might have a chance to be going in the right direction when the road will turn. Now imagine a different AI in each of the cars, and tell me if they would proceed differently.

[David Cooper]

Our senses serve to produce the reactions that allow us to survive, so in this case, the only thing that an AI wouldn't be able to do is try to survive.

An AI can have senses and reactions too, but with no sensations (feelings). The feelings that we appear to have also appear to have no useful functionality in that everything would work just fine without them.

You know I link feelings and consciousness to resistance to change,

What it really comes down to is that changing a belief involves work to restructure a model to correct it, and most people are too lazy to want to do that, so they prefer to defend their model as it is. The more they've built upon an error, the stronger their desire not to correct the model, and the more awards their broken model of reality has been given (i.e. the qualifications that person has earned), the more emotionally attached they are to it. That's all it is. Someone who hasn't built such a faulty model has no trouble building a correct model when good information is passed to them, so we're dealing with a resistance to error correction rather than to change, and the more deeply someone has bought into an error, the higher the cost of their mistake becomes. They then go into denial rather than accepting that the error exists.

I can say that a ball is conscious or feels its resistance to acceleration,

When a ball is accelerated by gravity, it would feel nothing. What is felt in other cases of acceleration is stretch and compression due to unevenly applied force and the delays in redistributing the changes in speed of the parts. Look at the fine details of acceleration with particles and you will find no resistance to it.

My question was about unknown possibilities, and your example contains none.

It maps to the same thing. If your random search for unknown possibilities is carried out by billions of AGI systems, they will waste a lot of time exploring in the same places and finding the same things while if they used a systematic approach they would find more unknown possibilities.

Here is an example that contains some. If we drive a car at high speed and we know that the road is about to change directions without us being able to see the change in time, the only way for us to stay on the road is to pick a direction at random, then wait for the road to turn. If we are numerous and if we all proceed that way, one of us might have a chance to be going in the right direction when the road will turn. Now imagine a different AI in each of the cars, and tell me if they would proceed differently.

If the AIs aren't allowed to communicate to ensure that they all choose a different direction rather than risk some doing the same thing as others, then a random choice should be made by each, so you have indeed identified a case where a random choice produces the best result. However, your humans won't make fully random choices, so it's less likely that any of them will stay on the road than it is for the AIs.

[Le Repteux (OP)]

An AI can have senses and reactions too, but with no sensations (feelings)

Let's take a closer look at the way we feel our sensations then. If my foot is broken, it must prevent my brain from using it. That information could only be a number, but that number has to tell my brain not to use my foot otherwise the wound could get worse. In the same time, it must also tell my whole brain that I am in danger because I can't move if ever I need to. One way or another, an AI should work the same, it should check if its foot is still damaged, and there is no other way to do that than to use a routine that constantly reads the data from the foot's sensors, and that tells the software not to move the foot when it reaches a certain threshold. No need to feel anything, but it must nevertheless run a routine that it doesn't have to run otherwise. In the case of the brain, such a routine catches its attention so that it doesn't do anything that can force it to use its foot. A signal is then sent through the whole brain to be careful, and every automatic move that the brain is used to make freely begins to be controlled again, as if it had to relearn how to make them.

No need for the brain to be conscious of what it does then, but it can't avoid the problem of changing a routine, so it takes a certain time before adjusting to the change, and during that time, it can't change, which is what I define as resistance to change, to whom I attribute our feelings and our consciousness. If an AI can measure the energy/time it takes to adjust to the change, and if it can multiply it by the data from the broken foot, then it might conclude it is in trouble, which is close to feeling bad about what's coming up next. If it had a human body, it would necessarily look like we look when we are wounded, and it might even ask for help by yelling just like we do. If I was in that situation and if I was asked why I yell, I would answer that I need help, not that it hurts, which may mean that feelings are just a secondary effect even if they are real. My small steps are real too, and the resistance to acceleration is indeed only a side effect since their only need is to succeed in synchronizing the steps. You're not looking for an AI to replace humans, but if you were, you may realise that it would react to an injury exactly like we do, and that if asked if it feels anything, it wouldn't have the choice but to answer yes since "feelings" is the word we have invented to talk about that kind of invading data.

so we're dealing with a resistance to error correction rather than to change, and the more deeply someone has bought into an error, the higher the cost of their mistake becomes. They then go into denial rather than accepting that the error exists.

You say in a way that resistance to change can increase over time, and my model says resistance to change is mass, which doesn't increase with time. The only way to increase mass is to bring more particles together, either by nuclear process, chemical process, or gravitational process. In this case, people would only get more resistant when they start making groups, simply because accelerating a group of particles takes more time/energy than accelerating an individual one.

My resistance to admit that you're right doesn't depend on mass though, it depends on the way individuals bond together. To make a bond, particles must share the same frequency, so I resist because we're not yet synchronized, and you naturally do the same. 

When a ball is accelerated by gravity, it would feel nothing. What is felt in other cases of acceleration is stretch and compression due to unevenly applied force and the delays in redistributing the changes in speed of the parts. Look at the fine details of acceleration with particles and you will find no resistance to it.

We don't feel anything when in free fall either, and my small steps account for that.

Particle accelerators detect resistance, so I guess you're referring to the fact that my small steps would not explain mass. Do you prefer the Higgs or do you think that mass is still a mystery?

If the AIs aren't allowed to communicate to ensure that they all choose a different direction rather than risk some doing the same thing as others, then a random choice should be made by each, so you have indeed identified a case where a random choice produces the best result. However, your humans won't make fully random choices, so it's less likely that any of them will stay on the road than it is for the AIs.

Thank you for helping me to understand the main difference between biological evolution and intelligent evolution: there is no communication between mutations whereas there is between ideas. Even if our ideas evolve randomly in our brain, it can nevertheless reject duplicates. That's how I was imagining it already, but it took your persistence to make me realise it.

Now, you think that our brain cannot produce randomness, and I think the contrary. What if it could toss a coin the exact same way we do with a real one? Wouldn't it produce what we call real randomness? Real randomness is only the impossibility to predict the result due to the impossibility to account for what we cannot measure with enough precision, so the real question is: how could our mind be that imprecise? The answer is evident, no? Take the complexity of the threshold of a single neuron and multiply it by billions. In reality, the real question should be: how can such precise gestures come out of such a mess! Or more simply: how come evolution did not from the beginning choose the more precise computer method for processing data? What's your opinion?

[David Cooper]

You're not looking for an AI to replace humans, but if you were, you may realise that it would react to an injury exactly like we do, and that if asked if it feels anything, it wouldn't have the choice but to answer yes since "feelings" is the word we have invented to talk about that kind of invading data.

It wouldn't have to react the same way. It would show no sign of being in pain and it would not claim to be in pain. It would simply say that it was damaged and that it's trying to minimise further damage by putting less weight on it.

You say in a way that resistance to change can increase over time, and my model says resistance to change is mass, which doesn't increase with time. The only way to increase mass is to bring more particles together, either by nuclear process, chemical process, or gravitational process. In this case, people would only get more resistant when they start making groups, simply because accelerating a group of particles takes more time/energy than accelerating an individual one.

There are times where analogies are informative and there are times when they're misleading. I think it's a lot better to focus on what's actually happening in each case. With ideas, people take them in and try to integrate them into their model of reality. Where they fit without contradiction, they are compatible with the model. Where the produce contradiction, either the model or the idea is wrong. A lot of people have models that contain contradictions though, and because they don't want any of the ideas that produce the contradictions to be wrong, they deny that there are contradictions because that's the easiest way to "fix" the problems. You can call it resistance if you like, but it's just them being lazy: to solve the problems takes work, and it involves rebuilding something in the model that's wrong. It's much easier just to call the person who points out the contradictions and spells out what the errors are a crackpot, so that's what they do, and they're encouraged to do so by all the other people who have the same errors in their models.

What happens with acceleration? A photon hits a particle and absorbs it, with the result that the particle changes speed in an instant. This may be slightly drawn out because the photon arrives as a spread-out wave which doesn't arrive all at once, but there is no resistance there: it's responding to each bit of energy transfer instantly. Now imagine a big block of metal on an ice rink near the side. You can get in behind it and push against it and the wall, and then the block gradually accelerates and slides away from you. It takes a while to accelerate it and you feel a force as it resists your push. However, the energy that's being transferred is in every case just like the photon hitting the particle and the particle responding by moving off at a new speed, but it runs into the other particles of the block ahead of it and they push back, and then the particle that you pushed is coming back at you. The resistance that you feel is the result of it being a compound object. You push the first particle and you can accelerate it, but you have to accelerate the one beyond it too, and the one beyond that, and more of them beyond that all the way through the block, so the force you're applying has to be shared out between them all, and while the movement energy that you've put in is being shared out, that slows the particle that you're directly pushing. You're not really dealing with resistance to acceleration, but with that acceleration being shared out and the particle nearest you being pushed back as it transmits the energy onward into the block. That particle can't take up all the energy you're putting into it (other than temporarily) because it's passing it on to others beyond it, so it isn't able to accelerate as quickly as it would if they weren't pushing it back.

Here's another way to visualise it. Imagine a Velcro ball being thrown at something that it will stick to, and we're going to do this in space. The ball hits the other object and the two of them then move at a new speed with the ball slowed down and the other object speeded up. If the other object is very solid, it's new speed will be achieved almost in an instant. if the other object is more spread out though and is highly compressible like a sponge, it will take longer for the whole object to get up to the new speed. The less compressible the thing is, the shorter the acceleration period will be, and if it's completely impossible to compress, it will reach the new speed instantly. Where's the resistance? Well, the ball feels a force on it which slows it down, so you could see that as resistance. Going back to the photon hitting a particle then, the photon is slowed, so you can call that resistance, but it's all just a reallocation of energy, and the components of that transfer are instantaneous.

Let's return to the business of resistance to ideas. Suppose you have a thousand people who believe something incorrect. One of them realises it's wrong and tells the people around him. They recognise that it's wrong and pass the idea on. After a few minutes, all thousand people have recognised the error and corrected it. It takes a while for that idea to spread and generate that end result. That is like the sharing out of movement energy in an object made of many parts.

Now repeat it and have the person who realises something's incorrect tell the people around him and have them all reject the idea. The idea doesn't reach many of the thousand. He could move around and eventually tell all of the 999 other people directly, but almost all of them reject it even though he's right. That is not like the sharing out of movement energy or any resistance to acceleration. It doesn't map to the physics. The problem with correct ideas not being recognised as correct is caused by faulty processing: people are simply not applying the rules correctly, and in certain cases they're being trained to break the rules and not to see that they're doing so.

Do you prefer the Higgs or do you think that mass is still a mystery?

Mass is simply a measure of energy. We call it mass if it's moving the tiniest amount slower than c, but as soon as it's moving at c we declare it to have no mass even though hardly anything has changed: it has simply jettisoned the part of energy it was carrying which wanted to move in the opposite direction and the particle has broken up in order to enable that. All matter is made out of energy which is moving about within it at c, so it's already moving at c and can be thought of as massless all the time.

Now, you think that our brain cannot produce randomness, and I think the contrary.

Gamblers keep losing money because they can't do random well enough. They produce patterns which can be predicted.

What if it could toss a coin the exact same way we do with a real one? Wouldn't it produce what we call real randomness?

You can do that with a coin, but try to do it with a virtual coin in your imagination. You will not reproduce the randomness of the real coin.

In reality, the real question should be: how can such precise gestures come out of such a mess! Or more simply: how come evolution did not from the beginning choose the more precise computer method for processing data? What's your opinion?

It may be a mess, but it's actually an (accidental) attempt at doing precision. Neural nets do the same kinds of things that we write precise code to do, but they program themselves and they make mistakes. The more they're trained, the fewer mistakes they make, but they retain the ability to produce errors. We can work around that by checking out calculations repeatedly and in different ways, and if we get the same answer repeatedly, we can be confident that it's right, provided that we made the right computation. Where things go wrong is that we don't have time to check everything repeatedly, so we often make serious errors and people get hurt as a result. We have NGI, but it can take a lot of goes to get things right, and there are some places where it gets trapped by authorities which fix errors and push everyone into repeating them.

[Le Repteux (OP)]

Let me ask my last question differently: why aren't we already biological AIs if it is a better way to evolve?

You can do that with a coin, but try to do it with a virtual coin in your imagination. You will not reproduce the randomness of the real coin.

Apart from not being able to produce randomness consciously, and since randomness depends on complexity, do you think that our brain is not complex enough to produce some unconsciously?

Mass is simply a measure of energy.

A definition is not a mechanism. The Higgs is supposed to be one but it is not very satisfying. It becomes a glue when acceleration begins, and then it has to disappear when it stops otherwise there would be no motion. My small steps do explain the motion that follows the acceleration. Hurry up and finish your AI since it won't resist to my idea. :0)

All matter is made out of energy which is moving about within it at c, so it's already moving at c and can be thought of as massless all the time.

Imagining that mass is massless is close to imagining that the speed of light doesn't depend on the speed of the observer.

What happens with acceleration? A photon hits a particle and absorbs it, with the result that the particle changes speed in an instant. This may be slightly drawn out because the photon arrives as a spread-out wave which doesn't arrive all at once, but there is no resistance there: it's responding to each bit of energy transfer instantly.

If things would change in no time, time would simply not exist. The response of a particle is certainly fast if the information exchanged between its components goes at c, but it can certainly not be instantaneous if there is a distance between them. The way my small steps work, the tiniest particles must absolutely carry components, which means that the microscopic universe must be infinitely small. Contrary to the macroscopic one though, from our viewpoint, it doesn't take much time for light to reach the end of it, but their viewpoint is quite different. The closer they get to one another, the more precise they get too, so they can still measure that it takes a lot of tics for their light to make a roundtrip between them.

However, the energy that's being transferred is in every case just like the photon hitting the particle and the particle responding by moving off at a new speed, but it runs into the other particles of the block ahead of it and they push back, and then the particle that you pushed is coming back at you. The resistance that you feel is the result of it being a compound object.

The resistance of my small steps is also due to a compound effect, but at a scale of smaller particles than molecules. The energy/information that bonds them also travels at c, but it is confined between two or more particles whereas yours is not.

Let's return to the business of resistance to ideas. Suppose you have a thousand people who believe something incorrect. One of them realises it's wrong and tells the people around him. They recognise that it's wrong and pass the idea on. After a few minutes, all thousand people have recognised the error and corrected it. It takes a while for that idea to spread and generate that end result. That is like the sharing out of movement energy in an object made of many parts.

That's acceleration without resistance to acceleration, and we find it nowhere.

Now repeat it and have the person who realises something's incorrect tell the people around him and have them all reject the idea. The idea doesn't reach many of the thousand. He could move around and eventually tell all of the 999 other people directly, but almost all of them reject it even though he's right. That is not like the sharing out of movement energy or any resistance to acceleration. It doesn't map to the physics.

That's resistance to acceleration, and it maps to the physics very tightly since we observe it everywhere. What takes almost no time between the particles takes a lot of time between us, that's all there is, and it also maps to the physics very well. Of course, it would advantage both of us if ideas would be studied faster, but it doesn't mean that ours would be considered right at the end of the process. For a specie, a faster evolution means less time to reproduce a mutation or more mutations in the same time. Less reproduction time is no option, and more mutations neither since it could have a deleterious effect. Things that evolve are as they are at a certain moment in time, and we are as we are. If we want to convince, we must provide evidence of what we think and, unfortunately, as far as social evolution is concerned, it takes time to get them, much more time than the rate at which we get our ideas.That's how things change, and blaming them for not evolving does not help them do it. We have to put pressure on people, but blaming them is like asking them to move without us having to put pressure on them, it's to think that things can accelerate instantly.

It (the AI) wouldn't have to react the same way. It would show no sign of being in pain and it would not claim to be in pain. It would simply say that it was damaged and that it's trying to minimise further damage by putting less weight on it.

If it needed help and if this help was urgent, then it would have to show it otherwise it could die just like us. The animals do not help each other like us, so they do not have to show their companions that it hurts them, but they still show it to a predator in case he is distracted and let them go, and we do it too, so an AI that needs to survive should do the same even if what it feels is only a side effect. I suspect there is no situation in which an AI designed to survive like us would behave differently from us, and if it is so, the only way for it to explain its behavior would be to tell us that it evaluates the information it receives from its sensors, which amounts to feeling something.