= 27d567bc2d48d9f40e9ccc20ea370b15)
NancyLebovitz comments on Rationality Quotes November 2013 - Less Wrong
You are viewing a comment permalink. View the original post to see all comments and the full post content.
= 783df68a0f980790206b9ea87794c5b6)
Loading…= b715d02e85f7b3b4ae65ca3d3e450868)
Subscribe to RSS Feed
Powered by Reddit
= f037147d6e6c911a85753b9abdedda8d)
You are viewing a comment permalink. View the original post to see all comments and the full post content.
Comments (388)
There is no escape from evolution (variation and selection).
Deliberate genetic selection is just more complicated evolution.
Sure there is. Organisms could, in theory, create perfect replicas without variation for selection to act on. Contrariwise, they could create new organisms depending on what they needed that would bear no relation to themselves and would not reproduce in kind (or at all).
If I could write an AI, the last thing I'd want is to make it reproduce with random variations. If I could genetically engineer myself or my children, I'd want to introduce deliberate changes and eliminate random ones. (Apart from some temporary exceptions like the random element in our current immune systems.)
I think you're overusing the term "evolution". If you let it include any kind of variation (deliberate design) and any kind of selection (deliberate intelligent selection), you can't make any predictions that would hold for all "evolving" systems.
In which theory? I don't think this is true if temperatures are above absolute zero, for example.
I suspect that you're being too restrictive- it doesn't seem like variation has to be blind, and selection done by replication, for 'evolution' to be meaningful. Now, blind biological evolution and engineering design evolution will look different, but it seems reasonable to see an underlying connection between them.
True, you can't create perfect physical copies or even keep a single object perfectly unchanged for long. But macro-scale systems designed to eliminate variance and not to let microscopic deviations affect their macro-scale behavior can, for practical purposes, be made unchanging. Especially given an intelligent self-repairing agent that fixes unavoidable damage over time.
So, what kind of statements are valid for all kinds of evolution?
The direction (and often magnitude) of expected change over time is generally predictable, for example.
Can you be more specific? What is the expected direction of change for all evolutionary processes?
In general, the entities undergoing evolution will look more like the complement of their environments as time goes on.
I'm sorry, I don't understand. What is the "complement of the environment"?
Suppose a gazelle lives in a savannah; we should expect the gazelles to digest savannah grass, flee from cheetahs, be sexy to other gazelles, etc., and become that way if not already so. I think Dawkins has a good explanation of this somewhere, but I was unable to find it quickly, that genes are in some sense records of the ancestral environment.
Similarly, internet memes are in some sense a record of the interests of internet users, and car designs a record of the interests of car buyers and designers, and so on. Is that a clearer presentation?
It seems clear what you mean (though not why you called it the complement of the environment). But I still don't see what's common to all kinds of evolutions, so maybe I'm still misunderstanding.
It's certainly true that any evolved object is a function of its environment and we can deduce features of the environment from looking at the object. But this is also true for any object that has a history of being influenced by its environment. A geologist looks at a stone and tells you how it was shaped by rain. An astronomer looks at a nebula and tells you how it was created by a supernova. "Being able to learn about a thing's past environment from looking at its present shape" is so general that you must have meant something more than that, but what?
What about AGI? Radical human enhancement? Computronium? Post-biological civilizations? All impossible?
Offhand, I think they'd all include variation and selection.
I've seen an argument that a nanotech organism with a reasonable level of error-correction could with high probability make error-free clones of itself until the heat death of the universe.
That assumes the lack of black swans. Not a very good assumption when we extrapolate things until the heat death of the universe.
True, but it would have to be an exceedingly black swan to result in evolutionary-like mutations rather than simple annihilation.
First, annihilation is good enough -- a destroyed nanobot fails at making "error-free clones of itself until the heat death of the universe".
Second, all you need to do is to screw up the error-correction mechanism, the rest will take care of itself naturally.
That seems plausible to me, but it's still likely to be subject to selection because of competition for resources. Depending on its intelligence level and ethical structure, it might also be affected by arguments that it should limit its reproduction.
The point is there'd be no variation for evolution to select on.
Why do you think there'd only be one sort of nanotech organism to select on and/or that perfect self-replication is the best or only strategy?
Well, the organism would need to be preprogrammed to survive in whatever environment it might find itself in until then.