There is no distinction between animate and inanimate, or between human and machine, at the most fundamental level. I expect that we will eventually create intelligent machines (in the unlikely event that our civilization evades the mass extinction our intelligence has created). But for now we have only the myth of “artificial intelligence.” That is all we will have until our engineers are able to define and measure intelligence.
The Field of “AI” Has No Definition of Intelligence
Alan Turing was an inventor of the computer, and he remains the most famous person in the history of the field known as “artificial intelligence” (which should never be written without quotation marks, unless the context makes the phrase explicitly aspirational). He wrote in an influential article that intelligence is too difficult to define, and that a definition is not necessary for creating intelligent machines (Turing, 1954, as discussed here) (Note that he wrote “intelligence” in his title but never in the main text. Instead he wrote “thinking” throughout the main text, implying that he equated the two).
In this article Turing presented what is said to be an “operational definition of intelligence.” This is now known as the “Turing Test.” It essentially says that if a thing appears to behave in an intelligent manner, it is intelligent. I am amazed that anyone takes this seriously, but they do. I see it, and its philosophy of functionalism, as one of many pathologies associated with our failure to use our theory of mind and our concept of observers (see The Neuron’s Point of View and Science’s Observation Problem).
Today’s engineers still have no serious definition of intelligence, let alone a serious mathematical measure of it. But many of them are nonetheless quick to tell us that ChatGPT and its progeny are highly intelligent, so much so that they will save us from ourselves. Or alternatively, transform us all into paperclips.
The Myth of “Artificial Intelligence”
A potentially greater and much less recognized threat is the widespread delusion that our machines are, or soon will be, at least as intelligent as humans. We have yet to make a machine worthy of being associated with the word “intelligent.” The world’s leading researchers are not on track to do so.
Intelligence is not to be found in what a thing does, but in what it is and what it knows. Common wisdom says that ChatGPT knows a tremendous amount, but this is not true, for reasons I will explain later.
For now I will just mention that what actually exists is “hardware,” and thus the reality of ChatGPT is only hardware. “Software” is a useful concept, but it does not exist as anything more than hardware. In itself it is only the rational product of our imagination, like the set of all real numbers (which are “real” only in the sense that they are less imaginary than the imaginary numbers). Software refers to algorithms, which are purely mathematical, and are defined only from the perspective of the software engineer. Science has a tendency to confuse mathematics with reality, in addition to its problem with observers and points of view.
Intelligence is Evidence
Rational Observer Theory says that intelligence, like everything else, is evidence. The more a thing knows, the more intelligent it is. I hope in the near future to publish a critique of “AI.” I will argue that our current machines do not have even the intelligence of worms.