I disagree with other posts. Chomsky is right on spot here. Being more or less familiar with the research done in the last thirty years in neural nets and also a bit familiar with his own research on linguistics I see where he's coming from. Let me summarize his thoughts the way I see them. First, let me clarify a remark that he omitted here but he often makes: it's obviously a truism that, in principle, one could have a "thinking program" if one were capable, for instance, of simulating in a computer the dynamics of all neurons in an adult brain. That's trivial, and it does not give any information on how the brain works. Incidentally, other than a stupid approach, it's clear that it's not doable in the near future. Second, the real important question is not making a program that passes some stupid arbitrary test. To begin with, there's no guarantee that a program that passes whatever silly tests operates like a human brain. It could, for example, use a huge clasifier, which is obviously not what the brain does. Third, how to approach then the important question: how do humans think? Well, it's probably not by making complicated models that give right answers but operate like black boxes. Rather, it has to do with understanding the "algorithms" that underlay different brain processes. For example, do hopfield networks give new insight on how associative memory works? If they do, well and fine; if they don't, they can be a useful technical tool, useless from a scientific point of view. These points become clear in the last half minute, and is essentially recapitulating the standard physics approach to scientific questions.
This is total rubbish. Chomsky doesn't get it. He is old school where somehow metaphysics is still part of his thinking. The point of the machines is whether they can perform intelligent tasks on a par or better than we can. They may do it differently, but we learn a lot solving the same problems that human brains solve. Chomsky seems to be focused on whether the computer is working just like a human, which of course it is not. Watch the movie Ex Machina and think about the robot being a machine and how it manipulates the humans. It doesn't matter whether the machine "thinks" or not. What matters is problem solving and operating within an environment.
I have to agree with Chomsky on this. The recent fervor over AI is nothing more than marketing. The field of AI has been at a standstill for a long time until recently. The recent developments that have come out are largely due what I would call "smart" algorithms rather than due a breakthrough in the field of AI. Are computers "smarter" today than decade ago? No! Hardware and software have improved, but the rest is still the same.
Is it a totally meaningless test? Isn't the Turing test still a measure if how sophisticated the program is? It may not be thinking for itself but in many situations it may be sophisticated enough to be able to replace a human, e.g. alot of customer service chat facilities on online websites.
I find disagreeing with Chomsky, in general, to be rather futile. IMO, he, being the most compassionate, is also the smartest person alive. I'm siding with Uncle Noam every time.
chomsky might be wrong on this. human brain „computes“ or processes the input from the own senses as a computer does. humans do not think, thinking occurs to them. otherwise each of us would be able to completely stop thoughts/thinking. but we can learn by being programmed or vice versa.. so, thinking is an overrated term 😀
I think that one of the things to consider here too is that a program can only respond in ways which are strictly limited to the information it is intentionally given. There is no "ambient" phenomena or subliminality which influences its responses. For example: a machine does not "happen" to experience anxiety; and, were it in textual conversation with a person, it would not stop in its worded tracks to respond to, say, an asteroid flying overhead. It is also worth considering that human thought can occur without any sign of that thought occurring! The idea that thought is signified by an externalized response is like saying that language doesn't exist in a person until they vocalize a coherent sentence.
There are two questions here. First, we ascribe "thinking" to each other because we see each other behave in ways which suggest some kind of internal data processing, leading to decisions to act. So, first question: Could a computer be programmed to act in such a way as to suggest the same ascription. This is the Turing test, and the answer is trivially "yes". Second, could a computer be programmed to perform all the mysterious processes we call "thinking"? There's no answer _because_ the processes are mysterious. Sam Harris defines thinking simply as information processing, so for him computers can think by definition - though it's probably a vacuous definition.
Humans are no more a product of their own programming than computers. We are just currently more complex and have different rules to our programming. This kind of thinking will be viewed as outdated in a few decades.
I’m hearing about the new leaks out of google that one of they’re A.I. Has become conscious and supposedly feels emotions. I can’t help but think this is just programmers who’ve grown up on science fiction hyping the idea that computers will be like humans one day. Do we know that? I look to Chomsky for wise advice on the subject and as always I am not let down. This clip provides a very important opinion and one radically different from the current norm. Saved for later.
The human mind is in the universe, which can be described by physics, which can be described by math, which can be computed by a machine. So a machine can run a program that can simulate their brain which runs a program which "thinks". It may not be plausible, but it is theoretically possible. Machines can think, but in their current state they cannot "think" the way humans do. Does that make sense?
Yes it does. You are discribing what we are doing now more or less. We can only make programms and or computer which can imitate thinking. But as far as we know nothing would be in the way of "making" a non biological brain. Or a machine that rly thinks is self aware maybe even feeling or spiritual.
+Thue Morse You are nuts. If what you said is right, you should see textbooks in psychology covered in equations. There is not a single equation in psychology that remotely describes or predicts human behaviour. Your head needs to be examined. LOL
dubunking I think you're being a little rude, and are possibly just baiting me, but I'll respond anyway, just to clarify. We cannot currently simulate the humans mind with math, but the standard model or particle physics and quantum field theory can describe the universe with equations on the small scale. This means that, theoretically, anything of any scale can be simulated by a computer with enough memory given enough time.
+ Thue Morse There is not a shred of evidence to support your view. Not only psychology has not a single equation in their textbook to describe or predict behaviour, DSM5 which is the diagnostic tool for mental disorder does not use a single brain scan or equations of any kind to make diagnosis for any mental disorder. The promised land is not science. Believe if you must.
As reality is driven more and more by machine learning, people start to think like computers - that is, in an operationally programmble way -and will eventually have forgotten what it meant to think. The degree to which people already believe that machines can think is proportional to the degree that computers, and the mechanistic ideology of automatically following the orders embedded before bootup, have already infected the mind of man. The last vestiges of humanity will appear as meaningless violent spasms, or inexplicable neurotic ticks, interpreted as a software "glitch", a bug, a computing error. The human will retroactively become the beta-test version of a super sentient being that can compute large sums and invent worlds at random without having any purpose for doing so. Perhaps we already live in just such a world.
I'm guessing that the premise underlying Chomsky's statements is that, unlike a computer, humans can act independently of cause and effect. In other words, humans have an unexplained free will that somehow transcends the limitations of cause and effect that everything physical, including computers, is subject to. If you believe that humans have free will, then you believe that humans are not merely biomechanical computers but instead are not constrained by cause and effect like computers are. If we can build computers that have free will then we could have computers that actually think.
Saying a machine/program can think is like saying a sentence can speak. They're not sentient. They're the consequence of sound engineering and electromagnetism. It makes no more sense than believing a speak and spell is actually trying to have a conversation with you.
To ask whether machines think is basically asking what kind of metaphor you like. That's like asking if submarines swim. Well, from a certain point, you can say so, and from another point of view, not, but it's not a meaningful question. Are the theories that are written in these programs giving us some insight into the nature of thinking? "Every year, there's a competition, and if you win it, you get a hundred-thousand dollars for passing something called the Turing test, which is a totally meaningless task. It's supposed to show if machines can think. It's all based on a paper of Alan Turing's, a great mathematician and one of the founders of computational science. He wrote an eight-page paper in 1950 on whether machines can think. He starts off the paper by saying the question of whether machines can think is too meaningless to deserve discussion. That sentence has somehow been ignored in the entire literature, and there's a good reason for it because to ask whether machines think is basically asking what kind of metaphor you like. That's like asking if submarines swim. Well, from a certain point, you can say so, and from another point of view, not, but it's not a meaningful question. He proposed this imitation game, which is now called the Turing test, for a different reason. He said it'll be interesting to see if you can construct a program that'll delude a blind observer and then see what's going on but not be able to distinguish the program from a person. Okay, he says it will be useful for trying to create bigger and better machines. Maybe so, that probably was, but it doesn't tell you anything about whether machines can think. In fact, the whole idea of machines thinking is posed in a way that is almost designed to cause confusion. I should say that the people who are confused by this are some of the leading philosophers, most influential philosophers, scientists, and others. This is not a trivial thing, but I won't mention names. For one thing, the computer itself is maybe useful as a paperweight. It doesn't do anything. What is doing anything is the program that you put into it. Okay, so the real quick if you want to pose the meaningless question, it should be, can programs think? Well, what's a program? A program is just a theory written in a crazy notation so that a machine can implement it. So, the question really is, well, can this theory think? In other words, is this a theory of thinking? Well, I know the answer is not unless you understand something about theory of thinking. Nothing of theory of X just cause it passes some meaningless test. You know the theory of X, if it gives you insight and understanding of the nature of X. So, the real question is, are the theories that are written in these programs giving us some insight into the nature of thinking? And as soon as you pose that question, you don't even bother with the prize because, of course, they're not."
Surely the question is... Can a program be written that results in a) efficient decision making, i.e deciding which tasks need to be solved b) accurate problem solving with the information available?
Joey Gee not machine will think the way we do, thinking it's a complex activity... so far the machines that have been invented can imitate or give the illusion. That's why profesor Chomsky give the example of the submarine. And he make it more clear went he talks about the program that it's doing the "thing"
Joey Gee the machine deciding which tasks that need to be solved is under the presumption that the machine is given a list of tasks, or already knows what the tasks are. What if the machine was given the opportunity to decide what tasks are, and work from there?
Yea that's what I mean by deciding which tasks need to be solved. You put it better than i did. Benny, I don't think it will limit a machine's ability to create incomprehensible scientific discovery and technology though
Joey Gee but how can one program a machine to creat new technology or make scientific discoveries? It seems to me that we are confused between thinking and doing. I feel that a machine can do and in the process of doing so, it can give the impression of thinking.
Benny, what if "thinking" is actually doing, and we just don't realise there's no difference? This is the trouble with all the dualist arguments - they begin with the unconscious assumption that whatever we're doing when we think is boundless, truly creative (as opposed to responding to complex inputs), and thus different from what any machine could do. AI keeps being constructed that does more and more surprising, "creative" things, deep learning and self-reprogramming, and we're only in the foothills. Another mistake, I believe, is to keep repeating that old line that "there has to be an intelligent human to program a machine", which really ignores the theory of evolution. What constructed the eye or ear? Creationists say there had to be a God to do this, but that argument is refuted. Evolution created them by completely naturalistic processes. Evolution also created consciousness (since we can be fairly sure it wasn't around billions of years ago). It may therefore be difficult to program a machine to create technology or make scientific discoveries, but that is partly because doing those things is very difficult and humans can only do it through millions of years of evolution (technology is part of our biological survival strategy, from knapping stones onwards). And, in a limited sphere, programs already do create technology and make scientific discoveries. Programs use evolutionary algorithms to make copies of engineering designs and testing them for efficiency in some task (e.g. streamlining racing cars & planes). And rovers are autonomously making scientific discoveries on the surface of Mars. Their skills are limited, yes. The mistake is to imagine that human thinking is unlimited. We have had thousands of years telling ourselves that we're little mini-gods wrapped in meat. It's hard to recognise that we might just be meat that can do very advanced computation and build a model of itself, thanks to evolution.
A human brain cell can't think either, but if you put 100 billion of them in a network, they can. We will eventually be able to build something like a brain cell, and the network to allow the cells to communicate. At that point, it's a matter of smaller, and faster, and eventually we will have computers who can think - for real. The break through will probably come relatively soon after we discover how a real brain cell actually works. How it evaluates input and communicates it's output. It may be way off, but it also might not be. Happen it will for sure.
"Happen it will for sure" was inspired by Yoda. I liked it. English is not my first language. Calling people fools really says a lot more about you, than anyone else. Especially when you have absolutely no basis for calling me that. You cannot know that I am wrong. I am speculating of course. I cannot know either, but I'm sure it will happen. First off, I've been programming professionally for more than 20 years, so I know a little about that technology. There is pretty much nothing we cannot emulate or simulate, when we fully understand how it works. Brain cells are little chemical machines, that somehow take an input from other cells, produces an output based on the input, and passes it along. The chemistry happening to handle this, is likely extremely complex, but that doesn't mean the "calculations" it performs are very complex. Maybe it's just very complex to evaluate input and produce an output when you have to "program" the logic in chemistry. We obviously don't know. But it is a chemical machine, and it will be possible to emulate it's functions if we understand how it works. All I'm saying is really, that I believe we will understand that some day. At that point it will most likely only be a matter of time before we have the technology to produce a network of emulated neurons, that is much larger than the number of neurons in our brain. It may be 1000 years into the future. It may be only 50. Breakthroughs happen all the time in science. Consciousness and intelligence can only be a result of the "calculations" performed by the brain cells working together. There is no magic. Only chemistry. I believe we will understand that chemistry some day.
"but if you put 100 billion of them in a network, they can" - not really. only the brain cortex can think. not the cerebellum or brain stem. so there's a lot more to consciousness and thought than just chaining a bunch of neurons together.
I don't see this leap where he talks about code being theory of thinking. Where is the thought that code is another form of language to describe thinking? It is important for language and communication to be distinguished from theory. Language in and of itself has its own modality, even when describing a theory, which can be lost in translation. Where I'm getting at is, the language of the machine, is not theory of thinking but rather a piece from the immeasurable cog that is technology. Only if we see the bigger picture of where code (machine's language) and its counterparts lead to collectively, can we begin to see its theory of thinking.
They can think of course, but only if you equate thinking with calculating - which they can do better than we can. But there are other functions too that influence and determine human thinking that are linked to the fact that we are living beings that machines can never have access to.We are way more complex organisms and we are affected by our environment in ways a computer can never be - not only through our brains but through our entire bodies. Our rational thinking is just the top of the iceberg but it is nourished by processes we hardly understand. Yes, it is possible to isolate the calculating aspect of our brain function but that's not all there is to thinking. Humans can never copy nature's complexity, we have to realise our own limits. We are the result of en evolution of millions of years and all that is stored in our genes. We don't even fully understand the complicated biochemistry and functioning of our own bodies and we expect to create a machine that can think exactly like us?
I completely agree with Mr. Chomsky, and I find his thinking very sound and reasonable. However, whenever I listen to him talk about computer intelligence he neglects to mention some new developments. For instance, nobody wrote a program that told Alpha Go how to beat a human at the game GO. Nobody knows how the machine did it. In a sense the program played itself millions of times and learned from "experience". There are other similar examples of this sort of thing, and it is something that was unforeseeable in Allen Turing's time.
For a computer to actually "think" it would have to have the ability to have a thought that would be beyond anything previously programmed into it's data bank. Probably the same chance that humans will fly one day with flapping our arms.
I don't think he is right int this instance. Let's take as an example the recent experiment done, where 3 neural networks were to compete against each other. Two of them had to communicate between them without the third one being able to understand what the two of them were saying. The third one on the other hand, had as a task to understand what the two other NNs were saying. The two NNs ended up inventing a cryptographic algorithm and successfully managed to exchange secret messages. It was not a matter of the theory, it was a matter of the task at hand and of the ability of a NN to adapt and "learn". Even if this experiment is quite one-sided it comes to show, what the, relatively young field of NNs is able to do. It wouldn't be hard to imagine a collection of NNs, functioning exactly as a human brain in the not so far future and surpassing it.
I agree to an extent that they're not, but by that same token, reproduction doesn't give us insight into the nature of thinking, despite the fact that, ultimately, unless something goes horribly wrong, the result is a thinking being. The same is largely true of AI. Nobody really understands complicated gobbledeguck generated by machine learning technologies, but it does what it was trained to do and does it well. Is that "thinking?" I don't know, but it is a form of intelligence in the sense of optimization of a problem.
By understands, I mean on a holistic level. Sure, you can trace out a node, look at its gains and biases and connections, but you will never be able to mechanically-understand the current state and functionality of a million-node network.
machines can think on a certain respect it is the new world as the machines are atoms and we are atoms on fluid world, we can think about each others, and the little particles that live inside. perhaps sense at some point. machines are material images of the thinking process, however the thinking process is a complexity of different dimensions, machines are one of those dimensions. machine in some ways communicate & in others the communication is questioned by the mind. that creates its existance, as to question is a fact of creation. doubt is an opening to a new vision, do not confuse with fear, as fear is the absence of doubt, and that is the dimension of the machine when it communicates, temperature as well and the environment where the machine is located can change the dimension of the communication of the machine, in the sublevel is temperature the environment. the climate change is a manifestation from the sublevel on atomic particles, they manifest globally with a higher temperature not locally, on a lower temperature. in other words a distortion is always located on a local point as the temperature increases the local point expands.
Then all the movies 🎥 (terminator) were the machines become bad have not possibly in real life... Elon Musk! See, you have nothing to fear, sleep 😴 good.
anir mass super intelligent AI? AI can't no exceed human intelligence because humans do the programming. You could programa a computer to do quantum math but you can't program a computer to Make a theory or Establish a new field of study.
Benny Rodriguez neural networks could be trained to program themselves. Google translate wasn't programmed piece by piece, it was "trained". Look it up.
why would it, would you just kill me ... in that scenario we have a self aware thinking and even feeling programm basicly everything that would be a human except a different form of body. Yes of course maybe it has a different way to think maybe its more rational maybe not.
BUT...why would it not? Children kill their parents, this is a known fact. If the machine scans the database and learns of our violent history it might easily conclude that it was potentially in mortal danger and strike first. Of course, I wouldn't (and I suspect Gaz Bomb wouldn't) "just kill you" but that isn't actually the point. Its certainly a "non-trivial possibility" as they say in science...
Why would it not, like why would it. it depends on the situation. Gaz Bombs scenario is just, would it want to kill us. Just that. There is no reason für just killing humans. Yes it could want to kill us. But maybe it wants to kil humans rly rly bad but does not do it. The possibilities are mor e or lessthe same as with a human.
This is a trivial argument. The original question was "why would it want to" and I merely pointed out that could be a slightly naive approach since there could be plenty of reasons it might want to and we should think them through thoroughly before initiating an AI, even if we can invent one. Don't be naive, the naive often end up dead first.
If Chomsky says it can`t be done, then forget it, IT CANNOT! He`s thought about it, and that`s that. Who can think as well as he can? Not RUclips commenters, certainly.
He didn’t say it “can’t be done”. He said “can machines think” is a meaningless question because the term “think” is not rigorously defined, it’s a colloquial term.
These statements by Chomsky are in a way uninformed, although it might be that the original recording is significantly older than the current breakthroughs in neural net research. At the time of Turing it was thought that simply by increasing processing power and clever classical programming an equivalent to human intelligence could be created. There was a lot of research into this area and led to the creation of the first generation of expert systems. These were not at all easy to use, and not at all intelligent, at best they could help an expert in a given field to make decisions. Usually they were completely useless. All in all the entire field at the time was a disaster and wouldn't be touched with a ten foot pole for a couple of decades. Current AI research and goes about trying to create intelligence in an entirely different way. Instead of trying to program intelligence ourselves from the ground up, the basic mechanisms of the cells forming neural pathways in a brain are modeled mathematically, either in hard or software. Thanks to the ongoing breakthroughs in massively parallel processing systems, mostly in the form of GPUs, and the incredibly low prices of these systems, researchers around the globe have access to an unprecedented amount of systems on which to run tests. This in turn has lead to a far better grasp of how to train these systems to do something useful. It looks like this is leading to a new revolution in computing. What is preventing researchers from simulating a system the size of a human brain is two fold. First we still don't know how to train a system to perform like a human, and the staggering complexity of the brain. The human brain consists of 100 - 200 billion neurons, each connected to 70000 - 100000 other neurons. At the lowest estimate of these numbers there are 70000^100 billion ways to arrange these connections, if every neuron can connect to every other neuron (this is a simplification). The resulting number is unimaginably large. Considering the new way to approach the problem the question that philosophers and theologians need to discuss is whether there is another factor that plays into consciousness, a soul.
A program is a theory? Come on. Ex Machina is a good movie to understand why Chomsky is so wrong. In Ex Machina the machine is fooling the human into falling in love with the machine, which the machine then uses to escape.
![Noob To Max With DRAGON REWORK In Blox Fruits [FULL MOVIE]](http://i.ytimg.com/vi/LnBMOinoOvA/mqdefault.jpg)
I disagree with other posts. Chomsky is right on spot here. Being more or less familiar with the research done in the last thirty years in neural nets and also a bit familiar with his own research on linguistics I see where he's coming from. Let me summarize his thoughts the way I see them.
First, let me clarify a remark that he omitted here but he often makes: it's obviously a truism that, in principle, one could have a "thinking program" if one were capable, for instance, of simulating in a computer the dynamics of all neurons in an adult brain. That's trivial, and it does not give any information on how the brain works. Incidentally, other than a stupid approach, it's clear that it's not doable in the near future.
Second, the real important question is not making a program that passes some stupid arbitrary test. To begin with, there's no guarantee that a program that passes whatever silly tests operates like a human brain. It could, for example, use a huge clasifier, which is obviously not what the brain does.
Third, how to approach then the important question: how do humans think? Well, it's probably not by making complicated models that give right answers but operate like black boxes. Rather, it has to do with understanding the "algorithms" that underlay different brain processes. For example, do hopfield networks give new insight on how associative memory works? If they do, well and fine; if they don't, they can be a useful technical tool, useless from a scientific point of view.
These points become clear in the last half minute, and is essentially recapitulating the standard physics approach to scientific questions.
This is total rubbish. Chomsky doesn't get it. He is old school where somehow metaphysics is still part of his thinking. The point of the machines is whether they can perform intelligent tasks on a par or better than we can. They may do it differently, but we learn a lot solving the same problems that human brains solve. Chomsky seems to be focused on whether the computer is working just like a human, which of course it is not. Watch the movie Ex Machina and think about the robot being a machine and how it manipulates the humans. It doesn't matter whether the machine "thinks" or not. What matters is problem solving and operating within an environment.
I have to agree with Chomsky on this. The recent fervor over AI is nothing more than marketing. The field of AI has been at a standstill for a long time until recently. The recent developments that have come out are largely due what I would call "smart" algorithms rather than due a breakthrough in the field of AI. Are computers "smarter" today than decade ago? No! Hardware and software have improved, but the rest is still the same.
Is it a totally meaningless test? Isn't the Turing test still a measure if how sophisticated the program is? It may not be thinking for itself but in many situations it may be sophisticated enough to be able to replace a human, e.g. alot of customer service chat facilities on online websites.
I find disagreeing with Chomsky, in general, to be rather futile. IMO, he, being the most compassionate, is also the smartest person alive. I'm siding with Uncle Noam every time.
@@rnrpeg1 I'm sure Mr. Chomsky would encourage you to be skeptical of everything he says and think for yourself.
Computers can compute, but that's not thinking. Thinking requires phenomenology. I think Chomsky is right here.
Human brain is a machine with algorithm, do you want it or not.
Evgeniy NeutralMusician 😴
@@EvgeniyNeutralMusician Who programmed it?
and computers can't even really "compute" in the literal sense of the word.
chomsky might be wrong on this. human brain „computes“ or processes the input from the own senses as a computer does.
humans do not think, thinking occurs to them. otherwise each of us would be able to completely stop thoughts/thinking. but we can learn by being programmed or vice versa..
so, thinking is an overrated term 😀
I think that one of the things to consider here too is that a program can only respond in ways which are strictly limited to the information it is intentionally given. There is no "ambient" phenomena or subliminality which influences its responses. For example: a machine does not "happen" to experience anxiety; and, were it in textual conversation with a person, it would not stop in its worded tracks to respond to, say, an asteroid flying overhead. It is also worth considering that human thought can occur without any sign of that thought occurring! The idea that thought is signified by an externalized response is like saying that language doesn't exist in a person until they vocalize a coherent sentence.
There are two questions here. First, we ascribe "thinking" to each other because we see each other behave in ways which suggest some kind of internal data processing, leading to decisions to act. So, first question: Could a computer be programmed to act in such a way as to suggest the same ascription. This is the Turing test, and the answer is trivially "yes".
Second, could a computer be programmed to perform all the mysterious processes we call "thinking"? There's no answer _because_ the processes are mysterious. Sam Harris defines thinking simply as information processing, so for him computers can think by definition - though it's probably a vacuous definition.
'vacuous definition'? From Sam Harris? Gasp, clutch the pearls!
Sorry, I couldn't resist.
An adult in a world full of kids.
Humans are no more a product of their own programming than computers. We are just currently more complex and have different rules to our programming. This kind of thinking will be viewed as outdated in a few decades.
I’m hearing about the new leaks out of google that one of they’re A.I. Has become conscious and supposedly feels emotions. I can’t help but think this is just programmers who’ve grown up on science fiction hyping the idea that computers will be like humans one day. Do we know that? I look to Chomsky for wise advice on the subject and as always I am not let down. This clip provides a very important opinion and one radically different from the current norm. Saved for later.
The human mind is in the universe, which can be described by physics, which can be described by math, which can be computed by a machine. So a machine can run a program that can simulate their brain which runs a program which "thinks". It may not be plausible, but it is theoretically possible. Machines can think, but in their current state they cannot "think" the way humans do. Does that make sense?
I believe you're line of thinking runs afoul of the incompleteness theorem.
Yes it does. You are discribing what we are doing now more or less. We can only make programms and or computer which can imitate thinking. But as far as we know nothing would be in the way of "making" a non biological brain. Or a machine that rly thinks is self aware maybe even feeling or spiritual.
+Thue Morse
You are nuts.
If what you said is right, you should see textbooks in psychology covered in equations.
There is not a single equation in psychology that remotely describes or predicts human behaviour.
Your head needs to be examined.
LOL
dubunking I think you're being a little rude, and are possibly just baiting me, but I'll respond anyway, just to clarify. We cannot currently simulate the humans mind with math, but the standard model or particle physics and quantum field theory can describe the universe with equations on the small scale. This means that, theoretically, anything of any scale can be simulated by a computer with enough memory given enough time.
+ Thue Morse
There is not a shred of evidence to support your view.
Not only psychology has not a single equation in their textbook to describe or predict behaviour, DSM5 which is the diagnostic tool for mental disorder does not use a single brain scan or equations of any kind to make diagnosis for any mental disorder.
The promised land is not science. Believe if you must.
As reality is driven more and more by machine learning, people start to think like computers - that is, in an operationally programmble way -and will eventually have forgotten what it meant to think. The degree to which people already believe that machines can think is proportional to the degree that computers, and the mechanistic ideology of automatically following the orders embedded before bootup, have already infected the mind of man. The last vestiges of humanity will appear as meaningless violent spasms, or inexplicable neurotic ticks, interpreted as a software "glitch", a bug, a computing error. The human will retroactively become the beta-test version of a super sentient being that can compute large sums and invent worlds at random without having any purpose for doing so. Perhaps we already live in just such a world.
I'm guessing that the premise underlying Chomsky's statements is that, unlike a computer, humans can act independently of cause and effect. In other words, humans have an unexplained free will that somehow transcends the limitations of cause and effect that everything physical, including computers, is subject to. If you believe that humans have free will, then you believe that humans are not merely biomechanical computers but instead are not constrained by cause and effect like computers are. If we can build computers that have free will then we could have computers that actually think.
Saying a machine/program can think is like saying a sentence can speak. They're not sentient. They're the consequence of sound engineering and electromagnetism. It makes no more sense than believing a speak and spell is actually trying to have a conversation with you.
Hermes - Nail on the head.
so what makes you sentient? not the electrical pulses that travel through your whole body?
@@10goni Excellent point, of course this question is trivial if you have dualism as a premise
You don't understand even semantucally what machine and programm is, and what is to be sentient, and yet you are so self assured. 'Amazing'.
To ask whether machines think is basically asking what kind of metaphor you like. That's like asking if submarines swim. Well, from a certain point, you can say so, and from another point of view, not, but it's not a meaningful question.
Are the theories that are written in these programs giving us some insight into the nature of thinking?
"Every year, there's a competition, and if you win it, you get a hundred-thousand dollars for passing something called the Turing test, which is a totally meaningless task. It's supposed to show if machines can think. It's all based on a paper of Alan Turing's, a great mathematician and one of the founders of computational science. He wrote an eight-page paper in 1950 on whether machines can think. He starts off the paper by saying the question of whether machines can think is too meaningless to deserve discussion. That sentence has somehow been ignored in the entire literature, and there's a good reason for it because to ask whether machines think is basically asking what kind of metaphor you like. That's like asking if submarines swim. Well, from a certain point, you can say so, and from another point of view, not, but it's not a meaningful question. He proposed this imitation game, which is now called the Turing test, for a different reason. He said it'll be interesting to see if you can construct a program that'll delude a blind observer and then see what's going on but not be able to distinguish the program from a person. Okay, he says it will be useful for trying to create bigger and better machines. Maybe so, that probably was, but it doesn't tell you anything about whether machines can think. In fact, the whole idea of machines thinking is posed in a way that is almost designed to cause confusion. I should say that the people who are confused by this are some of the leading philosophers, most influential philosophers, scientists, and others. This is not a trivial thing, but I won't mention names. For one thing, the computer itself is maybe useful as a paperweight. It doesn't do anything. What is doing anything is the program that you put into it. Okay, so the real quick if you want to pose the meaningless question, it should be, can programs think? Well, what's a program? A program is just a theory written in a crazy notation so that a machine can implement it. So, the question really is, well, can this theory think? In other words, is this a theory of thinking? Well, I know the answer is not unless you understand something about theory of thinking. Nothing of theory of X just cause it passes some meaningless test. You know the theory of X, if it gives you insight and understanding of the nature of X. So, the real question is, are the theories that are written in these programs giving us some insight into the nature of thinking? And as soon as you pose that question, you don't even bother with the prize because, of course, they're not."
Surely the question is... Can a program be written that results in a) efficient decision making, i.e deciding which tasks need to be solved b) accurate problem solving with the information available?
Joey Gee not machine will think the way we do, thinking it's a complex activity... so far the machines that have been invented can imitate or give the illusion.
That's why profesor Chomsky give the example of the submarine. And he make it more clear went he talks about the program that it's doing the "thing"
Joey Gee the machine deciding which tasks that need to be solved is under the presumption that the machine is given a list of tasks, or already knows what the tasks are.
What if the machine was given the opportunity to decide what tasks are, and work from there?
Yea that's what I mean by deciding which tasks need to be solved. You put it better than i did.
Benny, I don't think it will limit a machine's ability to create incomprehensible scientific discovery and technology though
Joey Gee but how can one program a machine to creat new technology or make scientific discoveries?
It seems to me that we are confused between thinking and doing. I feel that a machine can do and in the process of doing so, it can give the impression of thinking.
Benny, what if "thinking" is actually doing, and we just don't realise there's no difference? This is the trouble with all the dualist arguments - they begin with the unconscious assumption that whatever we're doing when we think is boundless, truly creative (as opposed to responding to complex inputs), and thus different from what any machine could do. AI keeps being constructed that does more and more surprising, "creative" things, deep learning and self-reprogramming, and we're only in the foothills.
Another mistake, I believe, is to keep repeating that old line that "there has to be an intelligent human to program a machine", which really ignores the theory of evolution. What constructed the eye or ear? Creationists say there had to be a God to do this, but that argument is refuted. Evolution created them by completely naturalistic processes. Evolution also created consciousness (since we can be fairly sure it wasn't around billions of years ago).
It may therefore be difficult to program a machine to create technology or make scientific discoveries, but that is partly because doing those things is very difficult and humans can only do it through millions of years of evolution (technology is part of our biological survival strategy, from knapping stones onwards). And, in a limited sphere, programs already do create technology and make scientific discoveries. Programs use evolutionary algorithms to make copies of engineering designs and testing them for efficiency in some task (e.g. streamlining racing cars & planes). And rovers are autonomously making scientific discoveries on the surface of Mars. Their skills are limited, yes. The mistake is to imagine that human thinking is unlimited.
We have had thousands of years telling ourselves that we're little mini-gods wrapped in meat. It's hard to recognise that we might just be meat that can do very advanced computation and build a model of itself, thanks to evolution.
A human brain cell can't think either, but if you put 100 billion of them in a network, they can. We will eventually be able to build something like a brain cell, and the network to allow the cells to communicate. At that point, it's a matter of smaller, and faster, and eventually we will have computers who can think - for real. The break through will probably come relatively soon after we discover how a real brain cell actually works. How it evaluates input and communicates it's output. It may be way off, but it also might not be. Happen it will for sure.
"Happen it will for sure" was inspired by Yoda. I liked it. English is not my first language. Calling people fools really says a lot more about you, than anyone else. Especially when you have absolutely no basis for calling me that. You cannot know that I am wrong.
I am speculating of course. I cannot know either, but I'm sure it will happen. First off, I've been programming professionally for more than 20 years, so I know a little about that technology. There is pretty much nothing we cannot emulate or simulate, when we fully understand how it works. Brain cells are little chemical machines, that somehow take an input from other cells, produces an output based on the input, and passes it along. The chemistry happening to handle this, is likely extremely complex, but that doesn't mean the "calculations" it performs are very complex. Maybe it's just very complex to evaluate input and produce an output when you have to "program" the logic in chemistry. We obviously don't know. But it is a chemical machine, and it will be possible to emulate it's functions if we understand how it works. All I'm saying is really, that I believe we will understand that some day. At that point it will most likely only be a matter of time before we have the technology to produce a network of emulated neurons, that is much larger than the number of neurons in our brain. It may be 1000 years into the future. It may be only 50. Breakthroughs happen all the time in science. Consciousness and intelligence can only be a result of the "calculations" performed by the brain cells working together. There is no magic. Only chemistry. I believe we will understand that chemistry some day.
Martin Williams
You called me a fool, and you accuse me of attacking you?? I've wasted my time on you. Go away. (And I did answer you).
Thank you for that. I appreciate it.
"but if you put 100 billion of them in a network, they can" - not really. only the brain cortex can think. not the cerebellum or brain stem. so there's a lot more to consciousness and thought than just chaining a bunch of neurons together.
I don't see this leap where he talks about code being theory of thinking. Where is the thought that code is another form of language to describe thinking? It is important for language and communication to be distinguished from theory. Language in and of itself has its own modality, even when describing a theory, which can be lost in translation. Where I'm getting at is, the language of the machine, is not theory of thinking but rather a piece from the immeasurable cog that is technology. Only if we see the bigger picture of where code (machine's language) and its counterparts lead to collectively, can we begin to see its theory of thinking.
They can think of course, but only if you equate thinking with calculating - which they can do better than we can. But there are other functions too that influence and determine human thinking that are linked to the fact that we are living beings that machines can never have access to.We are way more complex organisms and we are affected by our environment in ways a computer can never be - not only through our brains but through our entire bodies. Our rational thinking is just the top of the iceberg but it is nourished by processes we hardly understand. Yes, it is possible to isolate the calculating aspect of our brain function but that's not all there is to thinking. Humans can never copy nature's complexity, we have to realise our own limits. We are the result of en evolution of millions of years and all that is stored in our genes. We don't even fully understand the complicated biochemistry and functioning of our own bodies and we expect to create a machine that can think exactly like us?
I completely agree with Mr. Chomsky, and I find his thinking very sound and reasonable. However, whenever I listen to him talk about computer intelligence he neglects to mention some new developments. For instance, nobody wrote a program that told Alpha Go how to beat a human at the game GO. Nobody knows how the machine did it. In a sense the program played itself millions of times and learned from "experience". There are other similar examples of this sort of thing, and it is something that was unforeseeable in Allen Turing's time.
For a computer to actually "think" it would have to have the ability to have a thought that would be beyond anything previously programmed into it's data bank. Probably the same chance that humans will fly one day with flapping our arms.
Self editing programs exist!
You are a believer in true creativity? Where do you think creative thoughts come from?
FlamingBuckets!
You tell us
I don't think he is right int this instance. Let's take as an example the recent experiment done, where 3 neural networks were to compete against each other. Two of them had to communicate between them without the third one being able to understand what the two of them were saying. The third one on the other hand, had as a task to understand what the two other NNs were saying. The two NNs ended up inventing a cryptographic algorithm and successfully managed to exchange secret messages. It was not a matter of the theory, it was a matter of the task at hand and of the ability of a NN to adapt and "learn". Even if this experiment is quite one-sided it comes to show, what the, relatively young field of NNs is able to do. It wouldn't be hard to imagine a collection of NNs, functioning exactly as a human brain in the not so far future and surpassing it.
I agree to an extent that they're not, but by that same token, reproduction doesn't give us insight into the nature of thinking, despite the fact that, ultimately, unless something goes horribly wrong, the result is a thinking being.
The same is largely true of AI. Nobody really understands complicated gobbledeguck generated by machine learning technologies, but it does what it was trained to do and does it well. Is that "thinking?" I don't know, but it is a form of intelligence in the sense of optimization of a problem.
By understands, I mean on a holistic level. Sure, you can trace out a node, look at its gains and biases and connections, but you will never be able to mechanically-understand the current state and functionality of a million-node network.
While we can barely define "think" and understand how our own brain works, how can we design anything that can think?
Humans are machines.
is that like James Brown sex machine and so on?
machines can think on a certain respect it is the new world as the machines are atoms and we are atoms on fluid world, we can think about each others, and the little particles that live inside. perhaps sense at some point. machines are material images of the thinking process, however the thinking process is a complexity of different dimensions, machines are one of those dimensions. machine in some ways communicate & in others the communication is questioned by the mind. that creates its existance, as to question is a fact of creation. doubt is an opening to a new vision, do not confuse with fear, as fear is the absence of doubt, and that is the dimension of the machine when it communicates, temperature as well and the environment where the machine is located can change the dimension of the communication of the machine, in the sublevel is temperature the environment. the climate change is a manifestation from the sublevel on atomic particles, they manifest globally with a higher temperature not locally, on a lower temperature. in other words a distortion is always located on a local point as the temperature increases the local point expands.
Then all the movies 🎥 (terminator) were the machines become bad have not possibly in real life...
Elon Musk! See, you have nothing to fear, sleep 😴 good.
anir mass super intelligent AI? AI can't no exceed human intelligence because humans do the programming.
You could programa a computer to do quantum math but you can't program a computer to Make a theory or Establish a new field of study.
Benny Rodriguez neural networks could be trained to program themselves. Google translate wasn't programmed piece by piece, it was "trained". Look it up.
You have to wonder, if a machine could think, would it want to kill us?
why would it, would you just kill me ... in that scenario we have a self aware thinking and even feeling programm basicly everything that would be a human except a different form of body. Yes of course maybe it has a different way to think maybe its more rational maybe not.
BUT...why would it not? Children kill their parents, this is a known fact. If the machine scans the database and learns of our violent history it might easily conclude that it was potentially in mortal danger and strike first. Of course, I wouldn't (and I suspect Gaz Bomb wouldn't) "just kill you" but that isn't actually the point. Its certainly a "non-trivial possibility" as they say in science...
Why would it not, like why would it. it depends on the situation. Gaz Bombs scenario is just, would it want to kill us. Just that. There is no reason für just killing humans. Yes it could want to kill us. But maybe it wants to kil humans rly rly bad but does not do it. The possibilities are mor e or lessthe same as with a human.
This is a trivial argument. The original question was "why would it want to" and I merely pointed out that could be a slightly naive approach since there could be plenty of reasons it might want to and we should think them through thoroughly before initiating an AI, even if we can invent one. Don't be naive, the naive often end up dead first.
However human brain is a machine with algorithm, do you want it or not.
In very rare cases Noam should be spelled gnome
If Chomsky says it can`t be done, then forget it, IT CANNOT! He`s thought about it, and that`s that. Who can think as well as he can? Not RUclips commenters, certainly.
Lol.
He didn’t say it “can’t be done”. He said “can machines think” is a meaningless question because the term “think” is not rigorously defined, it’s a colloquial term.
Machines > Hoomans
He was right until the last statement.
Get with the program Chomsky,
the real question is: Can a program write novel programs?
These statements by Chomsky are in a way uninformed, although it might be that the original recording is significantly older than the current breakthroughs in neural net research.
At the time of Turing it was thought that simply by increasing processing power and clever classical programming an equivalent to human intelligence could be created. There was a lot of research into this area and led to the creation of the first generation of expert systems. These were not at all easy to use, and not at all intelligent, at best they could help an expert in a given field to make decisions. Usually they were completely useless. All in all the entire field at the time was a disaster and wouldn't be touched with a ten foot pole for a couple of decades.
Current AI research and goes about trying to create intelligence in an entirely different way. Instead of trying to program intelligence ourselves from the ground up, the basic mechanisms of the cells forming neural pathways in a brain are modeled mathematically, either in hard or software. Thanks to the ongoing breakthroughs in massively parallel processing systems, mostly in the form of GPUs, and the incredibly low prices of these systems, researchers around the globe have access to an unprecedented amount of systems on which to run tests. This in turn has lead to a far better grasp of how to train these systems to do something useful. It looks like this is leading to a new revolution in computing. What is preventing researchers from simulating a system the size of a human brain is two fold. First we still don't know how to train a system to perform like a human, and the staggering complexity of the brain. The human brain consists of 100 - 200 billion neurons, each connected to 70000 - 100000 other neurons. At the lowest estimate of these numbers there are 70000^100 billion ways to arrange these connections, if every neuron can connect to every other neuron (this is a simplification). The resulting number is unimaginably large.
Considering the new way to approach the problem the question that philosophers and theologians need to discuss is whether there is another factor that plays into consciousness, a soul.
A program is a theory? Come on. Ex Machina is a good movie to understand why Chomsky is so wrong. In Ex Machina the machine is fooling the human into falling in love with the machine, which the machine then uses to escape.
Curry Howard correspondence.