To Mike Tinter, 2008-09-07
In the constructal theory of pattern formation, all individual "agents" and nodes in the system are where the most pressing optimization problems are placed. This replaces a fractal view of reality with one that is more encompassing and related to optimization, constraints, growth, and so on. In your email, you are constraining it to *me*, or really any other reader that is reading your message, and you're expecting the novelty to bubble up and so on from the collective traces of memories and experiences that are 'captured' and 'cached' in the brains of the readers. Now, as we know, there's a genotype to brains and a genotype (in general) to their experiences and context ("ah, he's a professor, this explains things! (sort of)"), and you're using this in your example as a way to get creativity out of the system. This is no guarantee since you aren't building these 'culture systems' yourself - simply relying on the fact that they still exist. Okay, that's a little hard to explain, so I'll take another approach. Say that you were making an ai that was outside of the human realm of conexts in the galaxies and so could not write such an email expecting Bryan, or some other reader, to reply to it with some aspect of creativity - what then? Well, generally, it would have to build the novelty engine itself. It can't guarantee that novelty will be generated, this is true, but as long as it's building such engines it's able to optimize their design and see which ones produce more "novelty" (or near-novelty) than others, and this is where the idea of genetic algorithms come into play, and evolutionary processes as the only source of persistent novelty in the universe. Indeed, there's even evolution behind your readership with which you were making an example of bubble-up creativity, right? As long as you can build those "engines" (such as building an entire society, perhaps -- von Neumann probes and civilization-in-a-box (hi Bokov!)) in an increasingly competent manner, then you're doing as much as possible in aiming for that "guarantee". So, originally I replied with the genotype-phenotype problem in response to the AGI-creativity-issue that Mike brought up. Using the ideas I've expressed here, the way to get those types of evolutionary processes going is through the general architectures of selection and optimization, of using designs and prototypes to test out various genotypes and their respective phenotypes. Like I said earlier, the ability to "make a culture" or "make a civilization" (referencing Alex Future Bokov's old (and very dead) wiki, as well as others (civilization-as-a-creature from kk.org)) - and perhaps even to build brains [1], sort of like Greg Egan [2] [3]. On the extropy-chat mailing list a few days ago I mentioned that I have a prototype "brain on a disc", a very corny prototype at that, which is a linux live cd with computational neuroscience software packages on it. Eventually I'd like this to be "apt-get" for brain building and other manufacturing projects [4]. It's not enough, however, to just have "apt-get" and say you're done with it; you need neurofeedback systems (like those described in [2] or [5]) for competency improvement and improvement in the optimized strategies or else you're just going to end up with a static deathtrap that doesn't hook into our natural human ability to create and 'do stuff' (pardon my generalities). The other day I came across an interesting formulation of the computational basis of the mind, Mauk's modification to the "The Astonishing Hypothesis" -- "The astonishing computation hypothesis is that “You”, your joys, your sorrows, your memories and your ambitions, your sense of personal identity and free will, are in fact no more than different forms of computation implemented by a vast assembly of biology." And if it's just forms of computation, who cares whether it is implemented on biology or implemented on the cold touch of silicon (going both ways, mind you), or whether it's implemented particularly in your own forms of computation -- and maybe not just in your own brain but in a group of brains (brain farms, or just newly constructed cultures) -- thus you (quite literally **you**) have a place to begin and start to explore the realm of autocreativity. I'm not interested in folk psychology, in "goals", "ambition", "insight" or "truth", you must know that I'm just interested in fully functioning source code -- whether beginning in myself, in others or some collaborative extropic sense.

[1] http://heybryan.org/buildingbrains.html
[2] http://heybryan.org/humancortex.html
[3] http://heybryan.org/egan.html
[4] http://heybryan.org/exp.html
[5] http://heybryan.org/recursion.html

The above content could easily lead into an explanation of why I am working on building a 'fab lab' or 'fabuntu' and why I want to do closure engineering with SDKB to figure out the general design of self-replication, and why I want "fabuntu" to do physical manufacturing in my lab "right behind me". And why it is that I have been working on artificial ectogenetic tanks (artificial wombs) and DNA synthesizers and so on, but I'd rather go work on something else at this moment (bug me for more?).