Why "artificial intelligence" ("machine learning". "neural networks", etc.) is pure screw-over in a strictly mathematical sense.

Or, to be more precise, unscrupulous marketing designed to cheat incompetent investors, too much impressed in their childhood by pseudoscience fiction and films like A Space Odyssey and The Terminator.

Let us stage a simple mental experiment like Einstein's — but one that even a child can make.

Take a clock. An ordinary mechanical table alarm clock.

Is it "intelligent"? As "artificial intelligence' is rather a vague term (which is not accidental, as we’ll see below), we’ll specify our question: does a clock possess any consciousness?

Well, let’s imagine a far more complicated timepiece: a chiming clock with a tourbillon, perpetual calendar, and moon phases. And with dancing figurines. This is an exceptionally sophisticated mechanism, nearly the acme of watchmaking art.

The question is: How much closer is this clock to obtaining consciousness than our initial alarm clock was? The answer is obvious. The number of parts (gear-wheels, springs, balances and other mechanical odds and ends) in our latter clock will run into thousands.

Now imagine a brain-blowing super-sophisticated clock of a hundred billion parts (one part per every Chinese in one of a hundred Chinas – or twelve times the population of the whole Earth). FYI: this mega clock will be the size of a 17-storey building.

What are the chances of such a monstrous mechanism ever realize the tragedy of Prince Hamlet?

Alas. Yet our multi-billion-part hyper-clock has already reached the human brain’s order of complexity. Huh.

"Stop blowing smoke about clocks", you might say; for AI ("strong" one that is equivalent to human intelligence and possessing consciousness) will run on super cool computers!"

Alas again. It is a mathematical fact that any algorithm, however complex, can be reduced to a set of simple ones that can run on what is called a Turing machine. Mathematically, a clock and a super computer (even an incredibly powerful one) are essentially the same thing, a Turing machine.

Simply speaking, an algorithmic device.

The algorithms they execute may differ as widely as a baby’s fart and the relativity theory, but this is not important, for any complex algorithm inevitably falls into a set of elementary ones.

There is essentially no hindrance to reproducing the functionality of any sophisticated super computer in a simple (albeit large) set of mechanical parts. Yes, computers are not only electronic. Nor can they be digital only!

Electronic digital computers are simply the most cost-effective ones for now.

Attention, please.

If the apologists of "strong AI' told the public at large that they would add a couple of quadrillions of wheels to an alarm clock — to get a mechanical Shakespeare… How would the public at large react? Their feedback would be limited to pointing at their heads and asking, "Are you idiots?"

Yet, the common sense of 95 per cent of mankind is restricted to their everyday experience.

So when they are told about a similar prospect based on sand (a.k.a. silicon, to put it more handsomely), the public listens; for clockwork is something clear and familiar, while God knows what goes on in those microchips. Ignorance breeds piety. Mathematically, however, both options are quite the same nonsense.

All those promoted to the rank of Google programmers may keep making their algorithms as complicated as possible — till the cows come home, and giving them fancy names that hint at the machine’s ability to think (when 'neural networks' go out of fashion, some new marketing meme will appear), which won’t bring them a single step closer to human consciousness.

And all the money invested into the invention of this perpetual engine… "So you understand".

Summary.

Artificial consciousness CANNOT be obtained on today’s hardware for reasons just as fundamental as the reason why a rocket can’t travel faster than light.

Possibly (perhaps!) this can be done using a quantum computer. Something that they regularly promise to show us "soon".

True. The promises to produce a working thermonuclear reactor "in just a few years" have turned 70 already — and still remain promises.