Boosting the collective intelligence of machines

Stuart Russell recently expressed the idea that our brains are responsible for most of what we value:

So the way I think about it is, everything good that we have in our lives, that civilization consists of, is from our intelligence, it’s not the result of our long teeth or big scary claws.

This seems to conflict with the idea that culture is largely responsible for our success - an idea expressed as follows by Richard Dawkins in 1976:

Most of what is unusual about man can be summed up in one word: `culture'.

Indeed, culture may well lave led to the production of our large brains, according to the idea that big brains are meme nests.

Of course the ideas that brains and culture are our primary powers are not completely independent. Culture requires brains, and many animals have rudimentary cultures, so culture alone is not enough. Indeed even brains and culture are not enough. Whales have both in considerable abundance - but they lack opposable thumbs and so never invented technology.

If Stuart Russell is right then smarter machines are what we need. However if it is our collective intelligence that needs boosting, there there might be other, better ways of accomplishing this besides boosting the individual intelligence of machines. During the agricultural revolution, human development took off when humans crowded together in cities. At the same time, their levels of aggression and hostility went down and they became more sociable. It was networking - rather than individual intelligence - that was most obviously involved.

Machines are now also clustering together - in data centers and in the cloud. They also face barriers to communication and trade mirroring our own hostility, paranoia and distrust - in the form of firewalls and incompatible protocols. It is quite common for machines in adjacent racks to never communicate at all. Humans in skyscrapers are similarly anti-social, but this is hardly an ideal situation.

Theory suggests an obvious way of improving cooperation between machines: use cultural kinship. Shared memes result in cooperation in the same way that shared genes do. If we can (somehow) get the machines to share enough software they will start talking to each other more - and this is likely to accelerate the ongoing machine cultural explosion, mirroring - or rather extending - the human cultural explosion that kicked off for us thousands of years ago.

Technological determinism

Technological determinism is, broadly speaking the idea that technology drives history. It signifies the force of convergent evolution (particularly convergent cultural evolution).

In my article on that topic, I wrote:

Technological determinism provides a modern theoretical foundation for the progressive theories of evolution championed by Herbert Spencer.

Technological determinism contains the term "determinism" - which some interpret to mean: "completely determined". However obviously the future is only partly determined by technological forces - chance events play some role.

Kevin Kelly's recent (nice) book "What technology wants" is all about technological determinism - though it rarely mentions the term. Another book on the topic is Does Technology Drive History? The Dilemma of Technological Determinism.

Evolution is apparently getting at progressively better at crossing valleys between adaptive peaks - using techniques such as scaffolding and simulation to bridge the gaps. Unless the valleys are getting deeper and wider it would seem reasonable to expect technological determinism to become increasingly true over time - as civilization progresses.

As far as I can tell, this means an ever-closer approximation of universal instrumental values - and increasing abilities at dissipating free energy and generating entropy.

Unless things go wrong, that is. Some models of evolution permit outcomes which do not involve extinction - and yet do result in constrained and limited growth. These models typically involve one agent taking over the evolutionary process and then guiding the process according to its own whims. Such messed-up, short-sighted agents would presumably be assimilated by the first aliens they meet - but, if they are plausible, they could curtail evolutionary potential for an extended period of time. This situation - while it lasted would probably represent a failure of the thesis of technological determinism to make good predictions.

We don't really have much experience with such universal monopolies. However, we can see the negative effect that various monopolies have had in human history. However, some believe that we should create such a monopoly - to avoid the harmful effects of conflict - or to coordinate on a universal scale. For instance, here is Michael Anissimov on the topic. Apparently this is a desirable way to avoid burning the cosmic commons, via hardscrapple frontier folk.

Of course, such global monopolies are widely derided and denigrated as representing "the new world order" or "totalitarianism". Most seem to regard these outcomes as being undesirable. I share their views.

It is not clear whether such an outcome is consistent with evolutionary theory. Monopoly is like monoculture - and it quickly becomes the target of parasites. However, we can't yet be sure that such outcomes are completely implausible.