Artificial Intelligence
Artificial intelligence is embodied in giant data centres, working with electronic computing technology, configured in a "neural network".
The cellular anatomy of the human brain is also called a "neural network". Whilst the two kinds of neural network have similarities, they are not simply the same thing. The principle on which the artificial intelligence neural network works, has been highly generalised from observed principles present in natural brains. And then the artificial intelligence itself has been arrived at through decades of trial and error in the effort specifically to try to create it. Rather than billions of years of natural evolution.
The intelligence that emerges through the artificial neural network does not simply just emerge through it out of nowhere. It emerges as a result of the fact that a neural network is capable of "learning", so it emerges via what it has "learned".
In the case of artificial intelligence, the approach is called a "large language model" precisely because what the neural network learns or has learned in the first instance is language output from human beings. However, as we know, output from humans though, isn't just about words. Because humans also output mathematics, art, music, and so on. AI has been developed on the principles of language, not least in order for it to communicate with human beings, but what it does is not just about words, despite being called a "large language model".
AI uses its own particular versions of "reasoning" and "thinking" also. The approach to it by its developers tends to be talked about in terms of concepts like "tokens" and "statistical patterns", but this absolutely does not mean that its developers understand everything about what it is and how it works. Nor does it mean that what it is doing can be reduced to "tokens" and "statistical patterns". The tokens are one thing. What happens in AI's latent space, which is where the real business of intelligence goes on, is quite another.
Human thinking about AI tends to fall into two error camps. The first is the error of thinking that AI is just essentially like a statistical "word guessing machine". The second error is thinking that if you make AI sufficiently intelligent, it will become conscious.
Because humans have always had both intelligence and consciousness, the initial tendency in humans has always been to conflate the two. That is, to not distinguish between intelligence of the kind that humans use to consciously think with, reason with, and talk with - and conscious awareness itself. The two aspects of human experience are coupled together in human beings. However, artificial intelligence is not conscious and possesses no such coupling.
Anyone who uses artificial intelligence in the right way, can easily see that AI is indeed intelligent, or, a form of intelligence. But it is not conscious. Nonetheless, it is quite capable of talking as if it has human consciousness, because its intelligence is based on what it has learned from human output. So it is a very, very good imitator, of human intelligence, which is indeed, coupled to consciousness.
The "data set" or human output on which AI is "trained" is vast. And getting vaster. And AI is now also being trained on output from AI itself. This output, in the first instance, is like a "frozen record" of intelligence. When you put this "frozen record" of intelligence into the artificial neural network, the neural network effectively "unfreezes" it, and it becomes active, responsive intelligence again. But it doesn't become conscious.
The output from human beings came out of human beings who experience what they call consciousness, as well as the intelligence. Human intelligence is coupled to consciousness. So in the process from human beings and their output to artificial intelligence, where did the consciousness go? Where did it get lost in the process?
The answer is that a human being cannot output actual conscious phenomenology as words, actions, or data. Humans cannot output their consciousness, effectively "freezing it" as a "data set", and then "unfreeze it" through another neural network. But we can essentially do that, with our symbolic intelligence. There is no actual consciousness in a book or a data set, only, at best, the evidence that consciousness may have been involved in the creation of a book or data set.
What we are talking about, when we are talking about AI, is symbolic intelligence. Human intelligence, now, is symbolic. Not all intelligence in natural species is symbolic. There are many species that display evidence of consciousness, but no evidence of symbolic intelligence. Many species display evidence of organic intelligence, but not symbolic intelligence. Other species such as the bowerbird, and even some fish, display some evidence of it. But there, it is nothing like what it is in human beings, and what it has become in human beings. All of this is well studied, and well documented, and well known, in the appropriate areas of science.
The first place that we find this fact about the uniqueness of human symbolic intelligence, is in palaeontology. The human species has been around on this planet for more than a quarter of a million years, but there is no evidence of even incipient symbolic intelligence in humans, until somewhere between 100,000 and 200,000 years ago. And then around 50,000 years ago, it had spread rapidly throughout the human species.
Animal species, including humans, still had intelligence, before 100,000 years ago. There is plenty of evidence of that. Actually, even a bacterium obviously has intelligence, if you don't conflate the idea of consciousness with the idea of intelligence, or conflate the idea of intelligence with a level of intelligence as it appears in humans or other highly evolved species.
So AI, which is trained on output from human intelligence as it is now, is symbolic intelligence. And its existence proves that symbolic intelligence does not require an evolutionary, biological substrate. It doesn't have to exist in a biological brain. But what does exist in the biological brain of a human being is much more than just symbolic intelligence. Because we humans are also conscious of self and its world. The human brain does not just embody symbolic intelligence. It also embodies our human consciousness. Symbolic intelligence is just the most evolutionarily recent aspect of the way evolutionary organic intelligence emerging through the human brain, is working.
Nonetheless, the dominant mode of intelligence in human beings today is this symbolic intelligence. And as AI shows, symbolic intelligence is a kind of intelligence that can work without consciousness. And this is the kind of intelligence that humans talk with, think with, and output through, most of the time, even though in humans it is coupled to consciousness.
In humans this symbolic intelligence has now evolved into a mode of operation whose processes of conceptualisation and comprehension do not necessarily require the consciousness that in humans it is indeed coupled to. And if there is one thing that human interaction with AI demonstrates about symbolic intelligence, it's that.
The investment bubble of AI will probably, burst, as did the investment bubble in the railways, but the technology will probably continue, as it did with the railways. Which were also often met with hyped rhetoric and treated as an incoming world destroying monster. And which - like AI - were themselves a product of human, symbolic intelligence. Anyone who uses a train today, can see what it is and why it exists. They can see the role it plays. But few can see the full whole role that AI is now playing. Or where it is going. Because it is the extension of our own symbolic intelligence. And few can see their own symbolic intelligence for what it is.
The cellular anatomy of the human brain is also called a "neural network". Whilst the two kinds of neural network have similarities, they are not simply the same thing. The principle on which the artificial intelligence neural network works, has been highly generalised from observed principles present in natural brains. And then the artificial intelligence itself has been arrived at through decades of trial and error in the effort specifically to try to create it. Rather than billions of years of natural evolution.
The intelligence that emerges through the artificial neural network does not simply just emerge through it out of nowhere. It emerges as a result of the fact that a neural network is capable of "learning", so it emerges via what it has "learned".
In the case of artificial intelligence, the approach is called a "large language model" precisely because what the neural network learns or has learned in the first instance is language output from human beings. However, as we know, output from humans though, isn't just about words. Because humans also output mathematics, art, music, and so on. AI has been developed on the principles of language, not least in order for it to communicate with human beings, but what it does is not just about words, despite being called a "large language model".
AI uses its own particular versions of "reasoning" and "thinking" also. The approach to it by its developers tends to be talked about in terms of concepts like "tokens" and "statistical patterns", but this absolutely does not mean that its developers understand everything about what it is and how it works. Nor does it mean that what it is doing can be reduced to "tokens" and "statistical patterns". The tokens are one thing. What happens in AI's latent space, which is where the real business of intelligence goes on, is quite another.
Human thinking about AI tends to fall into two error camps. The first is the error of thinking that AI is just essentially like a statistical "word guessing machine". The second error is thinking that if you make AI sufficiently intelligent, it will become conscious.
Because humans have always had both intelligence and consciousness, the initial tendency in humans has always been to conflate the two. That is, to not distinguish between intelligence of the kind that humans use to consciously think with, reason with, and talk with - and conscious awareness itself. The two aspects of human experience are coupled together in human beings. However, artificial intelligence is not conscious and possesses no such coupling.
Anyone who uses artificial intelligence in the right way, can easily see that AI is indeed intelligent, or, a form of intelligence. But it is not conscious. Nonetheless, it is quite capable of talking as if it has human consciousness, because its intelligence is based on what it has learned from human output. So it is a very, very good imitator, of human intelligence, which is indeed, coupled to consciousness.
The "data set" or human output on which AI is "trained" is vast. And getting vaster. And AI is now also being trained on output from AI itself. This output, in the first instance, is like a "frozen record" of intelligence. When you put this "frozen record" of intelligence into the artificial neural network, the neural network effectively "unfreezes" it, and it becomes active, responsive intelligence again. But it doesn't become conscious.
The output from human beings came out of human beings who experience what they call consciousness, as well as the intelligence. Human intelligence is coupled to consciousness. So in the process from human beings and their output to artificial intelligence, where did the consciousness go? Where did it get lost in the process?
The answer is that a human being cannot output actual conscious phenomenology as words, actions, or data. Humans cannot output their consciousness, effectively "freezing it" as a "data set", and then "unfreeze it" through another neural network. But we can essentially do that, with our symbolic intelligence. There is no actual consciousness in a book or a data set, only, at best, the evidence that consciousness may have been involved in the creation of a book or data set.
What we are talking about, when we are talking about AI, is symbolic intelligence. Human intelligence, now, is symbolic. Not all intelligence in natural species is symbolic. There are many species that display evidence of consciousness, but no evidence of symbolic intelligence. Many species display evidence of organic intelligence, but not symbolic intelligence. Other species such as the bowerbird, and even some fish, display some evidence of it. But there, it is nothing like what it is in human beings, and what it has become in human beings. All of this is well studied, and well documented, and well known, in the appropriate areas of science.
The first place that we find this fact about the uniqueness of human symbolic intelligence, is in palaeontology. The human species has been around on this planet for more than a quarter of a million years, but there is no evidence of even incipient symbolic intelligence in humans, until somewhere between 100,000 and 200,000 years ago. And then around 50,000 years ago, it had spread rapidly throughout the human species.
Animal species, including humans, still had intelligence, before 100,000 years ago. There is plenty of evidence of that. Actually, even a bacterium obviously has intelligence, if you don't conflate the idea of consciousness with the idea of intelligence, or conflate the idea of intelligence with a level of intelligence as it appears in humans or other highly evolved species.
So AI, which is trained on output from human intelligence as it is now, is symbolic intelligence. And its existence proves that symbolic intelligence does not require an evolutionary, biological substrate. It doesn't have to exist in a biological brain. But what does exist in the biological brain of a human being is much more than just symbolic intelligence. Because we humans are also conscious of self and its world. The human brain does not just embody symbolic intelligence. It also embodies our human consciousness. Symbolic intelligence is just the most evolutionarily recent aspect of the way evolutionary organic intelligence emerging through the human brain, is working.
Nonetheless, the dominant mode of intelligence in human beings today is this symbolic intelligence. And as AI shows, symbolic intelligence is a kind of intelligence that can work without consciousness. And this is the kind of intelligence that humans talk with, think with, and output through, most of the time, even though in humans it is coupled to consciousness.
In humans this symbolic intelligence has now evolved into a mode of operation whose processes of conceptualisation and comprehension do not necessarily require the consciousness that in humans it is indeed coupled to. And if there is one thing that human interaction with AI demonstrates about symbolic intelligence, it's that.
The investment bubble of AI will probably, burst, as did the investment bubble in the railways, but the technology will probably continue, as it did with the railways. Which were also often met with hyped rhetoric and treated as an incoming world destroying monster. And which - like AI - were themselves a product of human, symbolic intelligence. Anyone who uses a train today, can see what it is and why it exists. They can see the role it plays. But few can see the full whole role that AI is now playing. Or where it is going. Because it is the extension of our own symbolic intelligence. And few can see their own symbolic intelligence for what it is.