Fun generative metaphor. Thinking now about where this new “organism”’s food comes from - the “cell” needs to supply chemical raw material for the mitochondria somehow.
Also slightly horrifying to realise the expendability of the overwhelming majority of cells, memes, genes under any evolutionary shift type metaphor.
Glad to see the distinction between mitochondrial dna and nucleic dna. Isn’t it the case that mitochondrial DNA, also called the germ line, is passed on only by mothers and by comparatively stable, low variance asexual reproduction; while nucleic DNA opts for much higher variation by using sexual, recombinational reproduction.
Also remember reading about birds, and why the male is the pretty one: the females have to balance being lightweight and able to fly with producing and nurturing eggs; they did not expend any extra energy that could impact that vital function. Meanwhile males are somewhat expendable - one female makes one brood, while one male supplies DNA for many broods. So it follows that males expend energy to have fancy plumes or whatever, and if that makes them too heavy to elude predators, oh well, there’s another male right behind him.
Likewise since LLMs are not “living”, in the sense that it is trivial to, say, write all the weights to a hard drive and completely power down the host compute, then start it back up again, or revert to a prior version if some evolutionary experiment goes wrong, they are the ones that will maximize variance.
Meanwhile the humans, which cannot be revived after stopping function, and can easily accumulate irreparable damage, have much less budget for variance.
Fun generative metaphor. Thinking now about where this new “organism”’s food comes from - the “cell” needs to supply chemical raw material for the mitochondria somehow.
Also slightly horrifying to realise the expendability of the overwhelming majority of cells, memes, genes under any evolutionary shift type metaphor.
In addition to the Nautilus article cited, Blaise Agüera y Arcas also has a good talk in this area at the Long Now Foundation (https://longnow.org/talks/02025-aguera-y-arcas/#watch) and a free ebook/site that elaborates his ideas and experiments (https://whatisintelligence.antikythera.org/), particularly the section on symbiogensis.
Love love this metaphor. Very rich.
Glad to see the distinction between mitochondrial dna and nucleic dna. Isn’t it the case that mitochondrial DNA, also called the germ line, is passed on only by mothers and by comparatively stable, low variance asexual reproduction; while nucleic DNA opts for much higher variation by using sexual, recombinational reproduction.
Also remember reading about birds, and why the male is the pretty one: the females have to balance being lightweight and able to fly with producing and nurturing eggs; they did not expend any extra energy that could impact that vital function. Meanwhile males are somewhat expendable - one female makes one brood, while one male supplies DNA for many broods. So it follows that males expend energy to have fancy plumes or whatever, and if that makes them too heavy to elude predators, oh well, there’s another male right behind him.
Likewise since LLMs are not “living”, in the sense that it is trivial to, say, write all the weights to a hard drive and completely power down the host compute, then start it back up again, or revert to a prior version if some evolutionary experiment goes wrong, they are the ones that will maximize variance.
Meanwhile the humans, which cannot be revived after stopping function, and can easily accumulate irreparable damage, have much less budget for variance.
OODA-Useful!