> I have gathered that induction of the rule behind a sequence of data > is > in general completely intractable, perhaps it is possible to show that if > the source of that data follows a rule that is constrained by locality, it > is perhaps practical? Could there be an efficient general agent for 3D > worlds?
3D worlds in general are just as uncomputable and intractable as linear sequences. The point is, rather, that our brains are built with biases corresponding to the PARTICULAR patterns in the PARTICULAR environments in which humans tend to exist. So we are efficient semi-specialized agents for the class of environments in which we've evolved. Because this class of environments is not all THAT narrow, we've evolved some truly general pattern-recognition and synthesis and enaction capability, coupled with a lot of more specialized capability of this nature. Much of the complexity of the brain has to do with the need to keep specialized and more generalized capabilities smoothly functioning together. > Interfacing an AGI to the real world seems like a good idea to me, the > real world has a huge amount of patterns in it, from the straightforward > (walking into walls hurts) to the extremely subtle, and a wide range > between > the two, and I am not sure if a good enough spread can be easily achieved > in > a virtual world - an AGI's development may stall due to the lack of a > suitable stepping stones, stepping stones that the real world provides in > ways that are perhaps very subtle? It's true that making a virtual world with the richness of the real world is damn tough. I agree that giving an AGI access to real-world sensors and actuators is important, yet I also believe that a lot of teaching and learning of AGI's can be done in simulated worlds. IMO, the practical hassles of robotics are larger as the benefits, for the current stage of AGI work. However, this is largely an issue of financing -- with bucks to hire a team of robotics engineers to solve the various niggling problems that inevitably arise when messing around with robots, the AGI-robotics option would probably seem a lot more appealing! -- Ben G ------- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/[EMAIL PROTECTED]
