Thread by @IntuitMachine

The greatest value of AGI research is that understanding general intelligence opens up an understanding of ourselves.

How is general intelligence research different from other cognitive science fields like neuroscience, cognitive psychology, cybernetics, artificial intelligence, and machine learning?

Although all these fields do tackle an aspect of general intelligence, they do not necessarily focus entirely on it. Rather they focus on many other narrow domains that may or may not involve general intelligence.

Different fields establish a narrow approach to inquiry that do lead to an incremental improvement of our knowledge about cognition. However, none of these fields exclusively focuses on the big picture. General intelligence research attempts to see the forest.

It is indeed intriguing that the word 'forest' derives from foreign and outside. It is as if it was always meant to be overlooked!

The 'artificial' part of AGI is meant to describe the method of scientific discovery. That of reverse engineering biology so as to create a synthetic model that demonstrates general intelligence traits.

This reverse engineering approach implies an acknowledgment that the parts are greater than the whole. That a reductionist approach favored by many fields such as neuroscience cannot ever explain general intelligence.

This synthetic recreation also implies the inadequacy of descriptive models that are conjured up by psychology and philosophy. It is not enough to describe what is observed. It is necessary that models are able to generate general cognitive behavior.

It's important to distinguish what the word 'general' means in the context of 'general intelligence'. Many have taken this to mean the same as universal intelligence. That is intelligence that can solve any problem.

This is *not* what I mean about general. The use of general or the g-factor in psychology is meant to describe an elusive human trait that measures how effectively a person solves problems. But it should not imply a universal ability for all problems.

Only the kinds of problems that humans have been known to solve. So there is a limit in the scope of what 'general intelligence' means. It's the kind of intelligence that reflects what humans can do and how they do it.

A computer can perform a massive amount of exact symbolic transformations. This is not what humans can do. How humans think is very different from how computers compute.

Human thinking isn't of course more similar to how other mammals think than how a computer computes. Of course, there is a cognitive bias of centuries of indoctrination to believe that animals do not reason.

This bias led to the primacy of logic and the false belief of GOFAI. If we can reason about our thought processes then surely this must be the way to understanding how we think. It is why the philosophy of mind exists!

Reality is strange enough that its essence is below our intuitive methods of reason. This is the same as how our brains work. They are strange things that evolved over billions of years.

Strange things require empirical methods to uncover. We could not have uncovered the strangeness of quantum mechanics without rigid experiments.

General intelligence is difficult to understand because it involves a strange process that we have yet to fully uncover.

We cannot understand general intelligence if we assume that it is intuitive and thus understandable by human minds. Rather, we must accept that how minds come about are indeed strange, and explaining why they are strange ironically tells us who we are.

So yes, a theory of general intelligence will not look intuitive. In fact, it will look very strange. So strange that if you haven't spent time searching, you would dismiss it at first glance!

@threadreaderapp unroll