AMMDI Blyg

Kevin Kelly says quite reasonably that we don't know what intelligence is yet, and he breaks it down into a bunch of different components. I have some quibbles with his framing:

A major part of our ignorance stems from our confusion about the general category of either electricity or intelligence. We tend to view both electricity and intelligence as coherent elemental forces along a single dimension: you either have more of it or less.

Quibble 1: There's been a live debate over the unity of intelligence for a long while. IQ theorists and Berkeley Rationalist types seem to think there is indeed a single dimension. Cognitive scientists tend to think otherwise. Marvin Minsky was very much in the multidimension camp. From his Society of Mind glossary:

Intelligence (7.1) A term frequently used to express the myth that some single entity or element is responsible for the quality of a person's ability to reason. I prefer to think of this word as representing not any particular power or phenomenon, but simply all the mental skills that, at any particular moment, we admire but don't yet understand.

Quibble 2 is that electricity did in fact turn out to be something pretty simple, that can be described in a few laws and put in an undergraduate textbook. Intelligence I don't think is that way (and neither did Minsky). In his view, thinking is not one thing, but a set of complex and intricate mechanisms and tricks.

We are using LLMs to answer questions or to code software without having a theory of intelligence. A real theory of intelligence is so lacking that we don’t know how our own minds work, let alone the synthetic ones we can now create.

The terrible truth about LLM-based AI is that we don't need a theory of intelligence. This is The Bitter Lesson. A coherent theory may never emerge, because thinking is just too complicated. There is no periodic table of simple components. Minsky's GOFAI was basically a failure, because it was too thoughtful, too concerned with understanding how minds work, and thus too slow to make progress, whereas the LLM approach is to not care about that, just pour in data and go brrr.