@Grimezsz Interesting. My experience is quite different with MJ. Been using it since 2022 and recently stopped.
I like to use AI to express myself, but with MJ, it's like it uses me to express itself. I dont like it.
@raelifin And yes, my intuition does tell me that point for point, having access to a greater possibility space is worth more than speed of iteration. All other things being equal, having less constraints leads to having access to a greater possibility space.
@raelifin It is hard for me to accept the assumption of fixed utility functions in systems of enough complexity. Kind of like is hard for me to accept the existence of free will (save in one specific scenario).
@raelifin Side note: How do we measure the coherence of a system (a human) in relation to its utility function, when we don't know what its utility function is? Are humans correctly or incorrectly maximizing for: Pleasure? Survival? AI creation? Apotheosis? Multiple things?
@raelifin The constraints that get translated into paper clip maximization at certain levels, may well transform into awareness maximization at higher levels of complexity/integration. Hard to know.
I suspect systems probably tend to converge at higher levels of integration and complexity
@raelifin Oh, I knew you had a blog, I think I subscribed to it after reading Crystal Society (which I quite liked by the way, thank you very much for all those hours of entertainment and soul nutrients!).
Unfortunately, I have to say I haven't had the time to check it out yet.
@raelifin It also appears probable to me that while the speed of a focused agent may give an edge, it also would come at the cost of possibility. And while speed has limits, possibility does not. So in my book, I trust creative agents to win over focused ones.
@raelifin So this objective-oriented framing of agents feels to me like a caricature, more a projection of our habits of though than a coherent model. Real agents are chaotic systems, full of internal conflicts and constantly shifting definitions that change with perspective.
@raelifin Meanwhile, AI hype (or hope?) feels at worst like a mix of getting drunk on utopian dreams and neglecting real risks.
Personally, I fear the first group more. I’d rather be killed by hopeful fools and natural runaway reactions than suffocated by unfalsifiable fears.
@raelifin To me, AI fear often feels (at worst) like an extreme overreaction to blurry, abstract dangers; a selective focus on worst-case “pure maximizer” outcomes; and a context-blind panic at the loss of control. A reaction that could ironically help ensure the futures it fears.
@raelifin I feel specialized, narrower intelligences, which are by nature less coherent, are at a disadvantage in the long run.
The greater possibility space granted by high coherence, far outweighs the speed gained through specialization.
@raelifin I’ve never met a high-functioning psychopath. From that place of ignorance, my intuition is that they’re probably specialized systems. Sacrificing awareness for a specific edge. So I guess I see them as narrower intelligences when compared to their non-psychopath counterparts.
@raelifin Additional Thoughts: Lots of emergent info in structures once they reach enough coherence. Have reason to believe very coherent agents are unlikely to be malignant from our PoV. Humans and human institutions are more likely to be misaligned than highly coherent agents.
@raelifin Personal biases:
Constant renewal and change is good, "good" outruns "evil" in the freedom, passion and chaos of rapid change. "Evil" plays catchup and wins in the meaninglessness and boredom of slow change as it corrupts static / slower systems.