I’ve read the posts upon group selection and decided this is kind of relevant. I don’t know math deeply, so this will be qualitative.
There are mutually beneficial symbioses—flowers + pollinators, roots + mycorrhiza, that kind of thing. They are robust not because they allow for altruism, but because they are mutually beneficial (usually). I think it is worthwhile to look into the meta-question of how to measure their efficiency, if AI ecology is comparable to non-intelligent things’ ecology. Is fitness of symbiosis equal to the sum of (maybe weighted) the partners’ fitnesses (which might be hard enough to measure)? What do you think? Is it time scale dependent? What if there are multiple AIs (and each of them does care about itself, but cannot directly eliminate any other, only ‘crowd it out’?)
Suppose there are two largely interdependent kinds of agents, F and P. F between themselves compete for resources, and so do P. F and P exchange resources at certain rates and are relatively ‘loyal’. (Though some P parasitize upon other P and so have only vicarious links with F). They promote each other’s survival, both short-term (bad season) and long-term (new generation), but a season fatally bad for P usually doesn’t eradicate F.
Their ‘loyalty’ is not absolute, e.g. P1 will only form liaisons with F1, but F1 can also work with P2-P45 simultaneously, though never with P46, and will gain most if it partners with P29. Most P can partner with most G, and vice versa. If an F has already negotiated treaties with several P, it can influence its readiness to make another commitment either way or not at all, but generally, both F and P leave more than enough offspring to start anew each season (iteration). But the iteration’s start and finish are different for each F and P.
Let P be humans and F—AI (originally these were plants and fungi.) How do you compare the efficiency of their partnership? I mean, we don’t need AIs that are not contributing to our welfare, right?
Sorry if this is an illegal analogy.
I’ve read the posts upon group selection and decided this is kind of relevant. I don’t know math deeply, so this will be qualitative.
There are mutually beneficial symbioses—flowers + pollinators, roots + mycorrhiza, that kind of thing. They are robust not because they allow for altruism, but because they are mutually beneficial (usually). I think it is worthwhile to look into the meta-question of how to measure their efficiency, if AI ecology is comparable to non-intelligent things’ ecology. Is fitness of symbiosis equal to the sum of (maybe weighted) the partners’ fitnesses (which might be hard enough to measure)? What do you think? Is it time scale dependent? What if there are multiple AIs (and each of them does care about itself, but cannot directly eliminate any other, only ‘crowd it out’?)
Suppose there are two largely interdependent kinds of agents, F and P. F between themselves compete for resources, and so do P. F and P exchange resources at certain rates and are relatively ‘loyal’. (Though some P parasitize upon other P and so have only vicarious links with F). They promote each other’s survival, both short-term (bad season) and long-term (new generation), but a season fatally bad for P usually doesn’t eradicate F. Their ‘loyalty’ is not absolute, e.g. P1 will only form liaisons with F1, but F1 can also work with P2-P45 simultaneously, though never with P46, and will gain most if it partners with P29. Most P can partner with most G, and vice versa. If an F has already negotiated treaties with several P, it can influence its readiness to make another commitment either way or not at all, but generally, both F and P leave more than enough offspring to start anew each season (iteration). But the iteration’s start and finish are different for each F and P.
Let P be humans and F—AI (originally these were plants and fungi.) How do you compare the efficiency of their partnership? I mean, we don’t need AIs that are not contributing to our welfare, right? Sorry if this is an illegal analogy.