A bit off-topic, but a thing I've always found puzzling is octopus intelligence. They aren't especially social (as far as we know, unless you count their arms as semi-independent beings), they don't really live all that long, and don't teach their young anything (they don't raise their young and die soon after mating).

Source: https://en.wikipedia.org/wiki/Octopus

I've heard some people claim that it's not necessarily the brain/body size ratio that determines intelligence, but the ability to manipulate the world around you. An octopus has 8 appendages, a dolphin has none (maybe its snout and tail could count as two, but are not nearly as efficient as an octupus's arm).

I've never seen any research that backs this up, but I'd love to read some more about these claims. If anyone cares to elaborate?

And you'd be surprised about how little we know about animal intelligence. You should look into the psychology professor Irene Pepperberg and her parrot Alex (Source: https://en.m.wikipedia.org/wiki/Irene_Pepperberg). Fascinating.

Arthur C. Clarke had a story -- "Second Dawn", if you want to look it up -- playing with the theme of intelligent beings who were unable to do much physical manipulation of the world around them.

Dolphins can communicate thereby changing the behavior of their peers. This could be viewed as a form of manipulation.

I'm personally not too surprised. The sea is a complex environment and octopuses are invertebrates in a mostly vertebrate-dominated world. They need to outsmart both their preys and predators.

Or does increased brain capacity promote cooperation?

That's a bit like debating what came first: the large beer or the voluminous glass?

Anyway, here's an interesting datapoint: there's something called "domestication syndrome", basically a set of traits that is common to diverse animals that have adapted to humans, from goldfish to cattle. One of those traits is reduced brain size. And the loss in brain size is more pronounced in social animals (dogs), compared to psychopathic loner species like cats. (No idea about goldfish. Really have to read up on goldfish sociology.)

The causality, per evolutionary theory, must obviously be "randomly larger brain" -> "better ability to cooperate" -> "increased survival". The other way around is Lamarckian Evolutions, which doesn't happen.

That's interesting, I've never heard of reduced brain size for domesticated animals.

Or perhaps navigating the complexities of social groups is a really hard task and requires a more developed brain.

So if we took the generative/genetic algorithms which invent languages, and we plug them (eliza/parry) into the ones which recognize cat videos, can we train the A.I. to be able to talk to each other about the models they are exploring in Kitteh?

Faster evolution.

The smartest peers mate with the females.