What are the historical analogues of AI?

Many people think that the historical analogues of AI are the important inventions of the past: the internet, perhaps, or the transistor, or the steam engine, the printing press, the invention of writing itself, or of stone tools.

However, AI is not at root an invention, but, as the name suggests, a new form of intelligence. There are many inventions that make AI possible – I don’t doubt that the latest Nvidia chips are produced by means of tools and processes at least some of which were invented, if not this year, then last year or the year before. Those are “inventions”. They are important, but they are not as exciting as AI itself – not by far – and there is a reason for that. The reason is that there is a big difference between the next marginal invention incorporated into a certain form of intelligence, and the arrival of a totally new kind of intelligence.

The historical analogues of AI, actually, are the past forms of intelligence that have dominated the planet. The best known and most often discussed contender for this list is human intelligence, which is usually thought to reside in the human brain (or nervous system), and to operate at the level of the human individual. This form of intelligence, actually, is not nearly so important as most people think, and was not nearly so dominant. If it was ever dominant then that dominance certainly ended in the cognitive revolution perhaps 100,000 years ago.

Most people think that the cognitive revolution was a time when human individuals became more intelligent, so you may be surprised by the preceding paragraph. I claimed that the cognitive revolution was the end of the reign of the human individual. What happened in the cognitive revolution seems not to have been that the human individual became more intelligent, but that the human society became more intelligent. We have good data on this. I won’t review it here, but the archaelogical proxies we have for the intelligence of human individuals show small changes during the cognitive revolution, while the archaelogical proxies we have for human societies showed huge changes during the cognitive revolution.

Societies are made up of individuals in the same way that individuals are made up of cells. Societies are knitted together by language in the same way that individuals are knitted together by chemical signalling. In an ultimate sense, you won’t find a real dividing line between the governing mechanics of a society and the governing mechanics of an individual. Similarly, if you carefully study an individual at the cellular level, you won’t find a fundamental distinction between the governing mechanics of a human and the governing mechanics of a cell. However, there really is a vast, vast difference between a human and a cell. It would be a terrible mistake to think of one of these as the “true” object to be analyzed in all correct discussions of intelligence. In the same way, there is a vast, vast difference between a human and a society, and it would be a major error to think of individuals (or societies) as where the intelligence “really” lies.

So the first analog for AI is actually the human society. A human society is an intelligent entity. It is, by most of the definitions used these days, a form of superintelligence. It is by far more powerful than human individuals, as evidenced by the fact that it is vastly more capable of aquiring resources and using them to reproduce itself. Imagine trying to stand as an individual – meaning without tools or collaborators – against a society in any kind of contest over resources.

If we go back before human societies existed, we do find a period in which the human individual was somewhat powerful, though it is actually far from clear that humans prior to the cognitive revolution were very powerful at all. The last real step change in intelligence prior to the cognitive revolution was probably the arrival of subjective intelligence itself, which we take as the ability to have thoughts and emotions that refer to sense experiences. We don’t know where exactly in history this came up, but we know that it is of itself a very powerful form of intelligence. The ability to have thoughts and emotions about your experience makes narrative possible. It means that you can live in a reality of your own construction, while gathering real resources and reproducing yourself with them.

Most people think of thoughts and emotions as fundamental. That is because most people are addicted to the form of intelligence constituted by thoughts and emotions. We are so accustomed to living within that form of intelligence that most people have never experienced anything that’s not mediated by thoughts and emotions. Due to that, they wrongly conclude that thoughts and emotions are just how things are, or perhaps are the very bedrock of intelligence itself.

However there is a more basic form of intelligence even than thoughts and emotions, and even this one is not any kind of bedrock reality, although most people think of it as such: it is the form of intelligence consistuted by objective experience, which is to say the five senses: seeing, hearing, smelling, tasting, touching. This one really tends to be thought of as reality itself, but it is not. It is more accurate to say that the thing we know of as “the physical world”, together with our experience of it, the two of which are not nearly so distinct from one another as most people think, is at root a form of intelligence designed to gain control of resources and reproduce itself using them.

I have explored four historical analogs for AI going backwards, roughly, in time. Putting them now into forwards historical order, they are:

1. Objective experience

2. Subjective experience

3. Intersubjective experience (human societies)

4. Modern AI

These are not the only four kinds of intelligence ever to have existed on this planet — far from it. They are not even the only four kinds of intelligence ever to have domainated the planet, though all four probably did dominate during a certain period. Each next kind certainly did not replace the prior kind or kinds — quite the opposite. In fact all four kinds of intelligence are present today.

In truth, all four of these forms of intelligence are artificial, in the sense that they make use of artifices (parts of the physical world) and produce, as part of their operation, an artifical (made up) world. For this reason, here at MAPLE we call these the Four AIs, and we emphasize that modern AI (the kind that uses neural networks with weights found by gradient descent performed by microchips made out of semiconductors and )