Except they don’t have eachothers’ sourcecode, so even inaction or pacifist policies incur non-infinitesimal risk of infinite loss.
I don’t see what sourcecode has to do with it. A Kif can (perhaps only) minimise the risk of infintite loss by keeping interaction to an absolute minimum, preferably zero. That is not mere “inaction”in the sense of standing quietly in a elevator full of other Kif. That is “Another Kif....RUN!!!”
Except they don’t have eachothers’ sourcecode, so even inaction or pacifist policies incur non-infinitesimal risk of infinite loss.
I don’t see what free riding has to do with it either.
Running from other Kifs also (but less obviously) increases risk of -inf sfik. Cooperating or living socially with other kifs vastly increases your chances of survival in the wild, I presume. You run into a risk equilibrium problem (which can be reduced to raw math if you have enough data about the world and precise sfik values for Kif life and thing-ownership, though it does become very complicated maths (i.e. far more complex than I can solve) due to instrumental value effects and relational comparisons of -inf sfik risks) pretty quickly, but it does seem analytically solvable.
So Kifs have strong incentive to work together in some manner in order to avoid the high risks of death when going at it alone in the wilds, but also various recursive functions that compute sfik with some instrumental -inf sfik parameters plugging into the +fin sfik actions. Since this is apparently the case for most Kifs working together, but you have no guarantee that all Kifs can reliably precommit to cooperation (which is where source code factors in—if you could read the source code, you could see which ones can reliably precommit and under what conditions and how to enforce this), so there’s also some incentive to freeride on the Kif cooperation wave and gain a few sfiks by killing some other kifs and taking some of their things, so long as your ability to do so and the marginal gains from this (as well as the weighted instrumental parameter of -inf sfik risk reduction) outweigh the increase in -inf sfik that your higher status incur.
As an intuitive guess, Kifs are extremely cautious socially, but on average power structures and hierarchies of exponential rarity still form, as every Kif is aware that other Kifs have incentive to freeride and this means a risk that they get killed in the process of this free riding.
Running from other Kifs also (but less obviously) increases risk of -inf sfik. Cooperating or living socially with other kifs vastly increases your chances of survival in the wild, I presume.
Vastly is infinitely less than infinity.
living socially
They don’t exactly sound like social animals to me. The ideas that Kifs need to associate with other Kifs, but risk infinitley
negative disutulity from doing are not obviously compatible, to put it mildly. They are cat-like solitary predators writ large.
Running from other Kifs also (but less obviously) increases risk of -inf sfik. Cooperating or living socially with other kifs vastly increases your chances of survival in the wild, I presume.
Vastly is infinitely less than infinity.
This isn’t what the math says. The math says:
Go alone: 35% chance of -inf sfik within the next year.
Cooperate with other Kifs: X% chance of -inf sfik within the next year, with X being a function of the number of Kifs in the group and the chance of freeriders and the chance of being selected and so on.
So clearly for some situations and some numbers, i.e. for some functions, cooperating is superior to soloing.
Hyperbolic discounting heuristics, i.e. not valuing the state and utility of their distant future selves as much as their immediate self, perhaps in some manner which implements an asymptotic returns system for sufficiently distant selves, I would wager.
Granted, that’s just the easiest explanation that comes to mind. You’re correct that since this isn’t stated (AFAIK?) it’s a curious issue.
I don’t see what sourcecode has to do with it. A Kif can (perhaps only) minimise the risk of infintite loss by keeping interaction to an absolute minimum, preferably zero. That is not mere “inaction”in the sense of standing quietly in a elevator full of other Kif. That is “Another Kif....RUN!!!”
I don’t see what free riding has to do with it either.
Running from other Kifs also (but less obviously) increases risk of -inf sfik. Cooperating or living socially with other kifs vastly increases your chances of survival in the wild, I presume. You run into a risk equilibrium problem (which can be reduced to raw math if you have enough data about the world and precise sfik values for Kif life and thing-ownership, though it does become very complicated maths (i.e. far more complex than I can solve) due to instrumental value effects and relational comparisons of -inf sfik risks) pretty quickly, but it does seem analytically solvable.
So Kifs have strong incentive to work together in some manner in order to avoid the high risks of death when going at it alone in the wilds, but also various recursive functions that compute sfik with some instrumental -inf sfik parameters plugging into the +fin sfik actions. Since this is apparently the case for most Kifs working together, but you have no guarantee that all Kifs can reliably precommit to cooperation (which is where source code factors in—if you could read the source code, you could see which ones can reliably precommit and under what conditions and how to enforce this), so there’s also some incentive to freeride on the Kif cooperation wave and gain a few sfiks by killing some other kifs and taking some of their things, so long as your ability to do so and the marginal gains from this (as well as the weighted instrumental parameter of -inf sfik risk reduction) outweigh the increase in -inf sfik that your higher status incur.
As an intuitive guess, Kifs are extremely cautious socially, but on average power structures and hierarchies of exponential rarity still form, as every Kif is aware that other Kifs have incentive to freeride and this means a risk that they get killed in the process of this free riding.
Vastly is infinitely less than infinity.
They don’t exactly sound like social animals to me. The ideas that Kifs need to associate with other Kifs, but risk infinitley negative disutulity from doing are not obviously compatible, to put it mildly. They are cat-like solitary predators writ large.
This isn’t what the math says. The math says:
Go alone: 35% chance of -inf sfik within the next year. Cooperate with other Kifs: X% chance of -inf sfik within the next year, with X being a function of the number of Kifs in the group and the chance of freeriders and the chance of being selected and so on.
So clearly for some situations and some numbers, i.e. for some functions, cooperating is superior to soloing.
Since all Kif end up dead, they all end up on minus infinity. The question then is why they ever bother doing anything particular inbetween.
Hyperbolic discounting heuristics, i.e. not valuing the state and utility of their distant future selves as much as their immediate self, perhaps in some manner which implements an asymptotic returns system for sufficiently distant selves, I would wager.
Granted, that’s just the easiest explanation that comes to mind. You’re correct that since this isn’t stated (AFAIK?) it’s a curious issue.