Intelligence and AI

[Yazata]

French AI guy Francois Chollet has his own definition of intelligence, as a conversion ratio: the efficiency with which one can convert the information that's available into the ability to deal with future situations.

He says,

"A consequence of intelligence being a conversion ratio is that it is bounded. You cannot do better than optimality -- perfect conversion of the information you have available into the ability to deal with future situations.

You often hear people talk about how future AI will be omnipotent since it will have "an IQ of 10,000" or something like that -- or about how machine intelligence could increase exponentially. This makes no sense. If you are very intelligent, then your bottleneck quickly becomes the speed at which you can collect new information, rather than your intelligence.

In fact, most scientific fields today are not bounded by the limits of human intelligence but by experimentation."

It's certainly interesting, but I'm not entirely convinced.

There are matters of speed and the fact that electronics is incomparably faster than human neurobiology. Even if a robot and I are both performing optimal conversions, just the ability to complete them in milliseconds is not to be dismissed.

There are issues of recognizing, accessing and knowing all the information that we might already have available. There's that ancient 'universe in a grain of sand' idea. If the universe is holographic somehow and all information about everything is already present everywhere in everything, then the optimal conversion rate might (just speculatively) potentially be infinite.

And there's a deeper and more difficult to define issue, I think. Can this definition of intelligence really account for the difference between a human and a dog? Perhaps we could argue 'yes', that the human in simply much better at converting past experience into the ability to deal with future situations. But that's uninformative unless we inquire into the details.

Can we really say that we know what optimality in data conversion consists in? Or how far away from it we currently are? The human evolution of language - and by means of language, abstract and conceptual thought - was something of a quantum leap that's totally inconceivable from the point of view of a dog. (To conceive of it would require precisely the conceptual thought that the dog lacks.)

So might there be other, better ways of thinking that are inconceivable to humans because we never evolved the ability to think in those ways? If there are, and if AI's acquire the ability to access them, humans might end up in the same relationship to the AI's that dogs are to us.

Optimality might be a lot further away than we imagine (or possibly can imagine).

[C C]

[...] So might there be other, better ways of thinking that are inconceivable to humans because we never evolved the ability to think in those ways? If there are, and if AI's acquire the ability to access them, humans might end up in the same relationship to the AI's that dogs are to us.

Optimality might be a lot further away than we imagine (or possibly can imagine).

Mentally illness or heavy insanity potentially illustrate alternative ways of thinking. But that weirdness is not always constructive and productive, and it can be contingently dangerous.

Certain kinds of insanity also demonstrate how randomness or chaotic choices and behavior are not free will (FW). The sane individual actually needs to conform to preferences and sensible routines (be predictable -- compatibilism); but also to be able to re-program oneself if those habits and traits have undesirable consequences (which belief in FW provides the ability to do). But that said, slight randomness doesn't undermine FW -- however, such is not necessary for FW, either.

Jack Vance (below) once wrote a science fiction story about the principles of reality altering for the Earth, so that the changes favored the way that the insane thought, with the latter thereby adapting and those who were rational being decimated. Bizarre AI might need a radical reality like that to prosper also. Probably not, though -- the weirdness would just destroy everything human in AI's path, as it replicated the nonsense onward.
- - - - - - - - - - -

THE MEN RETURN, by Jack Vance (1957)
https://www.gutenberg.org/cache/epub/715...mages.html

EXCERPT: Matters had not always been so. The Relict [sane human] retained a few tattered recollections of the old days, before system and logic had been rendered obsolete. Man had dominated Earth by virtue of a single assumption: that an effect could be traced to a cause, itself the effect of a previous cause.

Manipulation of this basic law yielded rich results; there seemed no need for any other tool or instrumentality. Man congratulated himself on his generalized structure. He could live on desert, on plain or ice, in forest or in city; Nature had not shaped him to a special environment.

He was unaware of his vulnerability. Logic was the special environment; the brain was the special tool.

Then came the terrible hour when Earth swam into a pocket of non-causality, and all the ordered tensions of cause-effect dissolved. The special tool was useless; it had no purchase on reality.

From the two billions of men, only a few survived—the mad. They were now the Organisms, lords of the era, their discords so exactly equivalent to the vagaries of the land as to constitute a peculiar wild wisdom. Or perhaps the disorganized matter of the world, loose from the old organization, was peculiarly sensitive to psycho-kinesis.

A handful of others, the Relicts, managed to exist, but only through a delicate set of circumstances. They were the ones most strongly charged with the old causal dynamic. It persisted sufficiently to control the metabolism of their bodies, but could extend no further. They were fast dying out, for sanity provided no leverage against the environment. Sometimes their own minds sputtered and jangled, and they would go raving and leaping out across the plain.

The Organisms observed with neither surprise nor curiosity; how could surprise exist? The mad Relict might pause by an Organism, and try to duplicate the creature's existence. The Organism ate a mouthful of plant; so did the Relict. The Organism rubbed his feet with crushed water; so did the Relict. Presently the Relict would die of poison or rent bowels or skin lesions, while the Organism relaxed in the dank black grass.

Or the Organism might seek to eat the Relict; and the Relict would run off in terror, unable to abide any part of the world—running, bounding, breasting the thick air; eyes wide, mouth open, calling and gasping until finally he floundered in a pool of black iron or blundered into a vacuum pocket, to bat around like a fly in a bottle.

The Relicts now numbered very few. Finn, he who crouched on the rock overlooking the plain, lived with four others. Two of these were old men and soon would die. Finn likewise would die unless he found food.

Out on the plain one of the Organisms, Alpha, sat down, caught a handful of air, a globe of blue liquid, a rock, kneaded them together, pulled the mixture like taffy, gave it a great heave. It uncoiled from his hand like rope.

The Relict crouched low. No telling what devilry would occur to the creature. He and all the rest of them—unpredictable! The Relict valued their flesh as food; but they also would eat him if opportunity offered. In the competition he was at a great disadvantage.

Their random acts baffled him. If, seeking to escape, he ran, the worst terror would begin. The direction he set his face was seldom the direction the varying frictions of the ground let him move. But the Organisms were as random and uncommitted as the environment, and the double set of vagaries sometimes compounded, sometimes canceled each other. In the latter case the Organisms might catch him....

It was inexplicable. But then, what was not? The word "explanation" had no meaning...