On Thu, Feb 10, 2005 at 05:08:05AM -0500, Brad Wyble wrote: > >Arguably, we're approaching the region where a very large, very expensive > >installation could, in theory, support a nontrivial AI. > > > Yes, *in theory*, but you still have to engineer it. That's the hard > part.
I agree very much. AI is hardware-limited. We need enough crunch to address the engineering/emergence part. This amount of crunch could very well be 10^6 of what's needed to run the result. > Let's do a very crude thought experiment, and for the moment not consider > evolving AI, because the hardware requirements for that are a bit silly. SWNTs are just molecules. You can make them by the ton. We're within touching distance of bulk molecular electronics. Current systems have somewhere 10^10 switches. A mole of switches is thirteen orders of magnitude more (and faster, too). > So imagine it like this, you've got your 10^6 CPU's and you want to make > an AI. You have to devote some percentage of those CPU's to "thinking" If I can order a 10^6 CPU box, so can everybody. It should be easy enough to recruit a few million users for a little AI @ home project (to search through the parameter space, not for a distributed mind). > (ie analyzing and representing information) and the remainder to > restricting that thinking to some useful task. No one would argue, I > hope, that it's useful to blindly analyze all available information. I'm not very interested how it's done. (Because I know I'm not smart enough to figure it out). I define a framework, and let the system fill in the blanks. The fitness function is easy enough to check. This is doable, barely. > The part that's directing your resources is the control architechture and > it requires meticulous engineering and difficult design decisions. > What percentage do you allocate? I don't. I have a physics simulator, and a population of ANN-controlled co-evolving critters. The ANN is just an automaton network, with none of interaction functions defined in advance (or preseeded with plausible defaults -- a spiking network with hundreds of types, and a *working* morphogenetic genome). > 5%? 20%? The more you spend, the more efficiently the remaining CPU > power is spent. There's got to be a point at which you achieve a maximum > efficiency for your blob of silicon. Two words: premature optimization. > The brain is thoroughly riddled with such control architechture, starting > at the retina and moving back, it's a constant process of throwing out > information and compressing what's left into a more compact form. That's > really all your brain is doing from the moment a photon hits your eye, > determining whether or not you should ignore that photon. And it is a > Very Hard problem. Yes, but it's a solved problem. Biology is rife with useful blueprints to seed your system with. The substrate is different, though, so some things are harder and others are easier, so you have to coevolve both. This is where you need to sink moles of crunch. -- Eugen* Leitl <a href="http://leitl.org";>leitl</a> ______________________________________________________________ ICBM: 48.07078, 11.61144 http://www.leitl.org 8B29F6BE: 099D 78BA 2FD3 B014 B08A 7779 75B0 2443 8B29 F6BE http://moleculardevices.org http://nanomachines.net ------- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/[EMAIL PROTECTED]
pgpCqZhmHQPSL.pgp
Description: PGP signature
