Some thoughts on A Theory of Everyone

I just finished reading A Theory of Everyone by M. Muthukrishna. Overall it is a quite enjoyable read, with some interesting ideas, and some less compelling ones. At the very least it is bound to get you thinking about some very relevant topics.

If from the title and blurb you, like I, were expecting something akin to psychohistory, the actual theory will be underwhelming. I did not hope for a full development as described in Asimov's Foundation, but the least I ask of a scientific theory is some predictive capability. I do not mean this as a critique, but rather as a mismatch of expectations.

Identifying energy availability, the conditioning of evolution, cooperation and culture as the main drivers of behaviour does strike a chord, and the reasoning is very well laid out, with examples from throughout history. But it just shows that this is an explanation that fits, without truly convincing that it must be.

One thing I do appreciate is how it brings forth some examples that can, if considered from a polarized view and stripped of the caveats, be out of bounds for the current zeitgeist. For instance, acknowledging differences in the distribution of traits across sex and race, with proper citations, or proposing a more nuanced discussion on DEI initiatives, arguing that they do not address the root causes of inequality and therefore cannot, on their own, solve the issue.

While I cannot help but agree that a further step in energy is needed, and that nuclear (fusion in the long term, and possibly fission as a bridge) is the most likely candidate, I find it hard to equate energy to other benefits as directly as the book does, especially when mediated through the social and cultural structures, whose extreme but mostly unnoticed influence it describes so convincingly.

I found the description of the effects of evolution and culture on our behaviour very informative and compelling, and discover some unexpected interactions linking culture to the environment in which the society develops.

The book, for me, decays towards the end, when it moves into proposing how to address the future. In essence, the main idea is to just do trial and error at a large scale, in a distributed fashion. And this is where I find the theory lacking the prediction capability that would at least help select more promising interventions to guide the search, instead of running a kind of cultural genetic algorithm with an ill-defined cost function on reality.

Finally, the final words on ML and AI sounded to vague and void, no more than unfounded optimism.

Still, I would recommend the read for all the new perspectives and interesting tidbits, and, if not for the theory itself, for its ability to pose thought-provoking questions.