In this, problems or projects of dissimilar type are mapped against potential outcomes, building up experience sets. As those projects exhibit behavior of success or failure, direction or attribute, they are then redirected. Nurtured, promoted, or killed, based on early trajectories of following prior behavior to predict outcome.

I think this type of modelling - taking prior experience or outcomes and making “course corrections” based on similarities to those attributes, most closely aligns to how (some of us) think. apologize for lack of familiarity with all the modelling techniques, but isn’t this a part of the puzzle? As brains, we don’t process “all the info” to make our decisions, we just match up similar experiences, whether driving a car, or playing chess, or walking…why should computers do all that? Because it can? Or are we following what it can do, vs what it should do?

Curious stuff though - thanks for musing!

Regarding your specific suggestions, it is indeed particularly fascinating to consider the different ways that a program might model itself. Isn’t that playing with matches though? :)

Yeah, I’m unable to stop thinking about this stuff, and the fact that there is no “clear answer” about the mysteries of mind actually just makes it more attractive — until I get tired and discouraged, then I guess I need a break for a while. It help that there are other folks like you out there who are also trying to think about these problems.

I will give your entry a careful reading later; I tend to get highly confused by implications of computability and similar abstractions, which unfairly biases me toward concluding that they are unimportant. At first glance, your ideas about convergence seem to be saying that there is a meaningful sense in which emergent macro-objects exist in an objective way, beyond the descriptive capabilities of observers — if a viewpoint of that type ends up being useful, I’d certainly rather not overlook it! If I have some thoughts about your approach I’ll continue on your blog. Unfortunately, most times I end up with questions about the meaning and applicability of terms, in order to take underlying intuitions about fundamental concepts and make them compatible enough for real understanding — which is an exhausting and involved process. It’s amazing how much each of us is like an island.

I’d stopped reading this blog, because you said you were done!

Nice post. It caused me to write up some ideas I have in a semi-related direction:

http://dragonlogic-ai.blogspot.com/2009/06/importance-of-uncomputable-models-i.html

I think your example of beauty is a particularly interesting case of emergence. As you say, it seems to be the machinery of our subconscious minds (which nobody yet understands as far as I know) that is primarily responsible for the conceptual models comprising “beautiful.” Whatever they are exactly, it’s really complicated and tied to lots of other mental machinery. I expect there is no useful approximation to it from other modelling methods (no math equations or statements in logic). Yet, despite the messiness it’s so useful and universal to our minds that we stick a short label on the concept and use it all the time.

For this reason I expect that poetry and other “nonrational” means are more effective than computer programming or other formal methods for helping further our understanding of these wonderful idiosyncratic aspects of humanity, and building mechanical minds with models of beauty that match ours in deep ways is likely to be quite difficult, I think. We might have to replicate the brain in considerable detail to achieve that.

As for practical uses, in the view I am expressing here, emergent is the same exact thing as modelable and although modelling some things (like beauty) in comprehensive and useful ways might be beyond our current grasp, there are lots of things that are not beyond our ability to model. In fact, that is one of the most practical and effective uses we have for computers. After I work through a couple other topics I’m planning to do a little essay on exactly that subject pretty soon.

As humans, we seem to always want to know how the universe “works”, especially in terms of how the human mind processes and perceives this “universe”

For me, where these intersect is the reality (outside our minds) and our perception of reality (within our conscience and sub-conscience minds). My personal belief is that our conscience mind can perceive simple systems quite well, but when they branch into a sufficiently complex system or network, we can no longer follow. This is where I feel our sub-conscious or intuitive minds take over and can perceive some of these complex systems.

Take for example “beautiful”
Can we accurately describe beautiful? not me
Can we consciously perceive beautiful? somewhat
Can we model it with computers? Fractals seem to approach this, but only superficially…
But, do we “KNOW” what it is when we perceive it? Yes
Can we agree that it’s “emergent” Probably…

So, what’s my point? Emergence seems (philosophically) to be a convergence of human subconscious thought and conscious perception.

I’m no professional philosopher, but I would like to know more about how it works…. more importantly, how I can perceive it better and get in touch with it.

To Derek:
What then is the practical use of modeling to model emergence?

I suppose it is possible that unique among all technological achievements in history, AGI will appear fully-formed with no useful intermediate step. To most people that sounds pretty dumb though.

SL4 is a list devoted to generally-shocking futurist issues. I’ll take your word for it that it’s the preferred venue for technical discussion of Friendliness theory. [IMO if the topic is as serious as its proponents think, some sort of non-wacko-infested place for its development does not seem unreasonable.]

[at any rate, I just wanted to mention Friendliness before abandoning it along with the rest of the fringe AGI-related elements]

It can’t somehow turn out that any intelligence is Friendly. It’s a solved problem: you can have an optimizer that goes either way, producing paperclips or recursive teddy bears.

The only thing the public needs to be shown is evidence of imminence, but no such case is ever attempted.

Evidence like what? End of the world is a futile lesson to learn.

I am completely perplexed that the believers don’t have some sort of forum for discussing approaches to solving the Friendliness problem, especially technical issues and underlying concepts.

SL4 list was such place (but you are aware of that, so what’s your question?), the problem is that this is a difficult technical question, very few people were able to even really understand the problem, much less contribute to the solution. Keep in mind that FAI a kind of AGI, problem people have been failing miserably for some time now, but in this form it’s only harder.