Here's [0] the conclusion of a paper on optimizing population sizes in genetic algorithms:
> The use of smaller populations result in lower accuracy of the solution, obtained for a smaller computational time. The further increase of the population size increases the accuracy of solution. This effect is observed to a population size of 100 chromosomes. The use of larger populations does not improve the solution accuracy and only increase the needed computational resources.
The same effect may be what this article is describing, applied to culture. As the audience grows, the neural resources applied to judgement grow, but after a point the accuracy does not. A larger population does not translate into a more thorough search of the solution space compared to the same population divided into smaller niches that do not communicate.
I doubt it would be a great improvement to divide the population back down to the tiny tribes of prehistory, but the optimum is likely somewhere between that and our global melting pot.
Inclusion is a modern creed and battle cry, but to a point exclusion also has its place in increasing innovation, in that it creates independent niches that each attempt to climb different maxima.
"The use of smaller populations result in lower accuracy of the solution, obtained for a smaller computational time."
Here's a different view:
"Small Populations over Many Generations can beat Large Populations over Few Generations in Genetic Programming"[1]
The results of AI research is often very problem-dependent, as there are so many variables that vary between problems, and many ways to tweak the algorithms themselves. They're also often not reproduceable. So I take their claims with a huge grain of salt.
Groups of primates grow up to only a certain rough number, depending on the species. After that communication problems, jealousies, etc, prevent the ones on top of the pack from being able to manage the situation.
As someone who barely uses social media, I don't really understand what someone means when they "communicate" with 60,000 people. I envision a guy standing on the concert stage with a megaphone and 60,000 below him.
I think there are more caveats than meet the eye. For example innovation on something massive like the nuke or original space flight took tens of thousands of people. But finding the best form on a phone is trivial enough that you need only a couple engineers designing but having many models around the world would be ideal for finding the best.
> The use of smaller populations result in lower accuracy of the solution, obtained for a smaller computational time. The further increase of the population size increases the accuracy of solution. This effect is observed to a population size of 100 chromosomes. The use of larger populations does not improve the solution accuracy and only increase the needed computational resources.
The same effect may be what this article is describing, applied to culture. As the audience grows, the neural resources applied to judgement grow, but after a point the accuracy does not. A larger population does not translate into a more thorough search of the solution space compared to the same population divided into smaller niches that do not communicate.
I doubt it would be a great improvement to divide the population back down to the tiny tribes of prehistory, but the optimum is likely somewhere between that and our global melting pot.
Inclusion is a modern creed and battle cry, but to a point exclusion also has its place in increasing innovation, in that it creates independent niches that each attempt to climb different maxima.
[0] https://annals-csis.org/proceedings/2013/pliks/167.pdf