Does TOE apply to robots?

[Scott Mayers]

Is that Theory of Everything? I would assume so.

No. Theory of evolution.

PhilX

Yes, evolution is always a function of nature from physics to chemistry to biology to any further forms.

My proposed definition of Evolution:

An X evolves if it's finite description, D, alters due to some environmental influence, M, to become some non-D through X,
AND
it persists in the same or its altered environment.

This is rather more technical but makes it general enough to apply to anything. An X that 'persists' only references its particular domain. So if this domain is just one person, say, then a person 'evolves' when some finite description of them alters due to some environmental influence, like a mutation or something that affects change, like learning, that alters it to be what it is in a different way to what it was, for good or bad. All that matters is that such a person still exists.

Normally we assume these changes as 'improvements' but it does not require this meaning beyond its own existence with respect to itself. One might, for instance, change due to 'reading' the Bible (a mutation), and thus affect them to be something new ....some non-old state in which they survive well in some strong church community, they moved into. But this does not mean this makes them evolve 'better' in some objective way.

The X, above can be extended to a species, as we normally define it as, rather than one person. Then X also means the extension of people birthed through (or caused by) them.

Therefore, in this definition, A.I. can evolve too. So far, A.I. has evolved in total dependence by us as an environmental factor. The question I'm guessing most may wonder, though, is whether such a particularly defined A.I. can eventually 'self-evolve' beyond a human environmental influence, as Leo would be thinking above. In principle, it could. But so far, all living things have always had some complex connection to multiple other evolving things in a sympathetic coevolutionary way. As such, we could technically do this ourselves but it would require a great deal of more effort on our part to devise it alone to get to the point of self-sufficiency without us.

[Obvious Leo]

Scott. As you're probably aware, evolutionary (learning) algorithms are already widely used in modern software systems but only as sub-networks embedded within an over-arching linear framework. I believe facial recognition systems are an example of this. According to my sister-in-law neural network computing is the next big thing and she is widely regarded as a leading expert in the field.

[Philosophy Explorer]

Scott. As you're probably aware, evolutionary (learning) algorithms are already widely used in modern software systems but only as sub-networks embedded within an over-arching linear framework. I believe facial recognition systems are an example of this. According to my sister-in-law neural network computing is the next big thing and she is widely regarded as a leading expert in the field.

To add comment to this, I've been seeing articles on neural network computing running for the past two months on Flipboard. I'll check around to see what I can find and add.

PhilX

[Obvious Leo]

Scott. As you're probably aware, evolutionary (learning) algorithms are already widely used in modern software systems but only as sub-networks embedded within an over-arching linear framework. I believe facial recognition systems are an example of this. According to my sister-in-law neural network computing is the next big thing and she is widely regarded as a leading expert in the field.

To add comment to this, I've been seeing articles on neural network computing running for the past two months on Flipboard. I'll check around to see what I can find and add.

PhilX

It's a fascinating but very complex field, Phil, and my interest is mainly in the philosophical aspects of it rather than in the technical detail, but it relates closely both to this OP and your other topic on quantum computing. The essential point about AI and evolving computers is that they are self-programming in a very real sense. They quite literally write their own software and when you think about it this is exactly what a mind does and this is what makes it a mind rather than an automaton.

[Arising_uk]

The out-producing have a better likelihood, but if not of them are fit, they won't survive to reproduce again, and if a single producing species is fit enough to continue to survive, the number of offspring becomes irrelevant.

But in the main it's about competition within the species that produces a new species? I take your point about fitness and survival but if the mutation is faster gestation and reproduction but with a shorter life - span then I think the new species will win out.

[Scott Mayers]

I was meaning to show that even if WE are the ones causing robots to 'evolve', we are just another type of environmental influence, like the mutation AND the environment that gives it value as existing when we use them. So robots still evolve. So that is why I extended this to the more specific question that I think we are asking: can a robot be evolved by us to a point it can later evolve without us. [Not necessarily absent of other non-human environmental factors.]

I thought of a cool kind of example I mentioned to my brother in conversation of this a few years ago. It goes as follows:

Imagine trying to devise a computer that could ideally function in some way AND be able to prevent itself from being turned off or malfunction (survive).

It would need some means to have peripherals, like arms/legs that could either move to seek new power sources should its power be exhausted or to say, assure it can alter something in its present environment to repair any potential disconnection to its source of power.

Then, its peripherals would have to be able to repair its own electronics. For this problem, it is more difficult because if it can repair itself, it must have a separated system in isolation to the part of the system it is repairing. Otherwise, if something depending on a single 'system' also operated the peripherals used to repair itself, it could be prevented from doing this effectively.

So not only would it require distinctly separate CPUs in some common system but have distinctly separate CPUs and system, like a separate computer altogether. But since even these two separate systems might equally be faulty, it requires a third that could continue and so on. The more separate units (not simply multicore CPUs or threads as we use now) they have, the better their survival as a collective. This is how our evolution has evolved us to be made up of distinct cells that can be operated in an isolated state as well as a whole. Each subsystem would have to have a minimum capacity to operate in isolation of the others.

This requires a meshed type network (or open variable route) between them. This may not be able to mimic life with actual hardwired electronics as these are too finitely limited. BUT, if it uses radio signals as a means to communicate between atomic subcomputers (like 'cell' phones), this could be done. Note that they don't have to actually be intimately connected like our own human cells. Ants act as a complex whole living entity as a colony in this way.

[Obvious Leo]

The out-producing have a better likelihood, but if not of them are fit, they won't survive to reproduce again, and if a single producing species is fit enough to continue to survive, the number of offspring becomes irrelevant.

But in the main it's about competition within the species that produces a new species? I take your point about fitness and survival but if the mutation is faster gestation and reproduction but with a shorter life - span then I think the new species will win out.

There's a lot more to evolution than the origin of species. Evolution is an information theory and in biological systems it has a lot more to to with gene adaptation and gene expression than it has to do with gene selection. These adaptive and expressive functions are also specified by environment which basically means that at a cellular level we are evolving all the time. That's why we die.

[Philosophy Explorer]

This article is dated 4/21/2015 which gives some history to AI. I'm searching for a good comprehensive article that'll go into more detail. In the meantime I think it's common knowledge that IBM is at the forefront of this exciting new field with Deep Blue and Watson and we also have Google and NASA. I'll keep on looking for articles:

http://www.wired.com/2015/04/jeff-dean/

PhilX

[Obvious Leo]

It's a fascinating and rapidly growing field, Phil, and my own interest in it is mainly from the point of view of philosophy. To get at the foundational principles which underpin the philosophy of computation it's useful to go back to the work of some of the earliest pioneers in the field. The first true information theorist was Leibniz but the major ideas of Leibniz were more fully developed by George Boole in his laws of thought and his treatises on the mathematical foundations of logic. It was Boole who formally elaborated Leibniz's idea that all true statements must be ultimately reducible to a temporally determined sequence of simple yes/no answers, a principle which eventually took us to the binary logic gate and the computer chip. In the 2Oth century it was Alan Turing who took this to the next level but Turing was no philosopher. To really understand what Turing managed to achieve we need such geniuses as John von Neumann, Claude Shannon and John Conway. It was mainly these guys who laid down the underpinning principles which lie behind the self-programming computer but it was Turing who most simply defined the universe as such a thing, even though he didn't realise it himself. The Universal Turing Machine is the eternal reality maker which programmes its own input and thus can never make the same reality twice. He didn't know it but he could have lifted this idea straight from the mind of Heraclitus, or even from the Upanishads.

[Philosophy Explorer]

I was meaning to show that even if WE are the ones causing robots to 'evolve', we are just another type of environmental influence, like the mutation AND the environment that gives it value as existing when we use them. So robots still evolve. So that is why I extended this to the more specific question that I think we are asking: can a robot be evolved by us to a point it can later evolve without us. [Not necessarily absent of other non-human environmental factors.]

I thought of a cool kind of example I mentioned to my brother in conversation of this a few years ago. It goes as follows:

Imagine trying to devise a computer that could ideally function in some way AND be able to prevent itself from being turned off or malfunction (survive).

It would need some means to have peripherals, like arms/legs that could either move to seek new power sources should its power be exhausted or to say, assure it can alter something in its present environment to repair any potential disconnection to its source of power.

Then, its peripherals would have to be able to repair its own electronics. For this problem, it is more difficult because if it can repair itself, it must have a separated system in isolation to the part of the system it is repairing. Otherwise, if something depending on a single 'system' also operated the peripherals used to repair itself, it could be prevented from doing this effectively.

So not only would it require distinctly separate CPUs in some common system but have distinctly separate CPUs and system, like a separate computer altogether. But since even these two separate systems might equally be faulty, it requires a third that could continue and so on. The more separate units (not simply multicore CPUs or threads as we use now) they have, the better their survival as a collective. This is how our evolution has evolved us to be made up of distinct cells that can be operated in an isolated state as well as a whole. Each subsystem would have to have a minimum capacity to operate in isolation of the others.

This requires a meshed type network (or open variable route) between them. This may not be able to mimic life with actual hardwired electronics as these are too finitely limited. BUT, if it uses radio signals as a means to communicate between atomic subcomputers (like 'cell' phones), this could be done. Note that they don't have to actually be intimately connected like our own human cells. Ants act as a complex whole living entity as a colony in this way.

"...can a robot be evolved by us to a point it can later evolve without us...." For months it's been a theme in many articles that they would be a threat to mankind which strongly implies they can. For many it's a question as to not if, but when? I don't think it"s inevitable that robots will be a threat, but I do believe they will have the capacity to evolve if they don't have it yet. This article goes into details:

http://www.theguardian.com/technology/2 ... en-hawking

PhilX

[Scott Mayers]

"...can a robot be evolved by us to a point it can later evolve without us...." For months it's been a theme in many articles that they would be a threat to mankind which strongly implies they can. For many it's a question as to not if, but when? I don't think it"s inevitable that robots will be a threat, but I do believe they will have the capacity to evolve if they don't have it yet. This article goes into details:

http://www.theguardian.com/technology/2 ... en-hawking

PhilX

I personally would not be 'sad' (nor 'happy' either, of course) should robots evolve to replace us regardless. Even if we were to evolve long enough to persist in another 100,000 years, we'd likely look back on our 'humanity' in this time much like we perceive Neanderthals. It doesn't matter what or which evolves in the future unless we were able to be there and judge one way or the other. A comet or asteroid will likely come along at some point and destroy whatever being running the show at that time.

[Philosophy Explorer]

"...can a robot be evolved by us to a point it can later evolve without us...." For months it's been a theme in many articles that they would be a threat to mankind which strongly implies they can. For many it's a question as to not if, but when? I don't think it"s inevitable that robots will be a threat, but I do believe they will have the capacity to evolve if they don't have it yet. This article goes into details:

http://www.theguardian.com/technology/2 ... en-hawking

PhilX

I personally would not be 'sad' (nor 'happy' either, of course) should robots evolve to replace us regardless. Even if we were to evolve long enough to persist in another 100,000 years, we'd likely look back on our 'humanity' in this time much like we perceive Neanderthals. It doesn't matter what or which evolves in the future unless we were able to be there and judge one way or the other. A comet or asteroid will likely come along at some point and destroy whatever being running the show at that time.

My concern is the loss of jobs and sales. It doesn't wash with me that new jobs will come along because robots can fill them too at some point as the technology improves. Of course people won't have spending power which raises further problems for businesses. This situation doesn't seem to concern the authorities as they don't talk about it.

It's my understanding that one of the things that keep robots from taking over is they still can't use their hands in the manner of humans. That's one of the things to keep an eye on (judging from the videos, depending on the situation, that may change in the next five to ten years). These are some of my thoughts.

PhilX

[Obvious Leo]

It's my understanding that one of the things that keep robots from taking over is they still can't use their hands in the manner of humans.

This is only one of many things, Phil, but the major one is the extremely primitive state of machine learning. The human brain has 100 billion neurons connected by up to 100 trillion synapses, many of which fire in parallel in cognitive processes and each of which can fire on a wide range of different action potentials. Some clever nerd worked out that this configuration means that the human brain has more logic gates than there are ATOMS in the entire universe. I don't know whether he was right or not but it seems that the robots have a long way to go. Currently the only fully autonomous artificial neural network has 11 neurons connected by about a 100 synapses. They have one on the drawing board which will have 100 neurons connected by about 1000 synapses and this is expected to be operational within a decade. If successful this algorithm would then have the learning capacities of a nematode worm. We need to be careful in a conversation such as this to make a distinction between an ability to store and retrieve information and an ability to create new information, which is what learning is.

[Philosophy Explorer]

This article explains how machines can learn:

http://bcs.mit.edu/news-events/news/mac ... arn-people

PhilX

[Obvious Leo]

This article explains how machines can learn:

http://bcs.mit.edu/news-events/news/mac ... arn-people

PhilX

The article gives a nice simple overview of the prodigious task the AI geeks have in front of them. Cognitive neuroscience is still in its infancy when it comes to the electro-chemical mechanics of learning because it wasn't until the the development of MRI imaging that much in the way of meaningful information could be acquired. There is currently an international project under way called The Human Connectome Project which aims to map every single synaptic pathway in the human brain. It'll be decades before this project is completed but the computing power alone which will be needed to achieve this is breathtaking. My sister-in-law is a leading theorist in learning algorithms and currently these are very primitive pieces of code which are inserted into ordinary linear computing code, so it is true to say that there are already computers which have a limited capacity for learning which isn't just number-crunching masquerading as learning. However the consensus in the field seems to be that true AI is still a long way off and the prospect of a computer ever being as intelligent as a human being is strictly for the SF writers. Certainly if such a thing were possible one would need to be certifiably insane to actually build such a thing because minds have a habit of doing whatever they fucking well like.