On Wed, Feb 2, 2011 at 9:53 AM, Colin Hales <c.ha...@pgrad.unimelb.edu.au> wrote:
> In relation to Stathis' request: > > If you model a natural environment presenting some problem to a human > within that environment, the simulated human will arrive at the same > solution as the real human would have. If intelligence is > problem-solving behaviour, there is therefore no difference between > the natural world and the model provided that the model is in fact a > good one. Your claim that computers cannot replicate human > intelligence is thus equivalent to a claim that there is some process > in the human brain which is not Turing emulable. What process do you > > > No. This is just plain wrong. You cannot model an observation of something > that you have no idea of the evidence of .i.e. You cannot model the unknown. > If you could then you'd already know it (the observer and the relationship > of the observer to everything else. If you want to get at unknowns, then you > have to model a modeller of the unknown ... and then _assume_ that > everything in a model captures the reality you are modelling, during the > process. Of course you can model the unknown: there would be no point to computer models otherwise. You include in the model the rules determining the system's behaviour, run it, and prepare to be surprised. > The non-Turing emulable part of the natural world is the relationship > between every little bit X and every other bit of it that is NOT directly > related to X. A product of massive parallelism created by a massive > collection of the entities of which we are actually made, which is best > assumed not to be abstract numbers if you want to understand it. This is > something we inherit by 'being' in the world. Something that cannot be > simulated. Something that a Turing Machine (computer), totally different to > us physically, does not get in its program. > By way of example, I have attached a video of a simulated neuron firing. > It's from a paper I have in review at the moment. The video depicts the > currents originating the biologically realistic EM fields around a neuron > due to the ion channels involed in an action potential. It was produce by > the package NEURON. In it you will see a pair of red/blue interfaces > travelling away from the soma. These interfaces are virtual evanescent > current-dipoles. They are mathematically describable, but form no part of > the mathematical description that generated them. THAT is what is missing. > These are the virtual relationships not accessed by the mathematics of a > Turing machine. No matter what is going on in a Turing machine, NONE of this > kind of phenomenon are accessed by it. If the NEURON model is good enough then you can use it to simulate a collection of neurons such as the human brain. This would allow you to simulate the motor output of a brain when it is presented with sensory input: in other words, when your model is presented with a problem it will process the data and present you with a solution, the same solution as a biological brain would have. The proviso is, of course, that the model is good enough. Do you think that there is some aspect of neuronal biochemistry that cannot be modelled at all by a computer? > The question is 'what is it like to BE those fields'. It cannot be claimed > to be like the mathematical description that represents them, nor can it be > claimed to be 'like' being the computer running the simulation. We can agree that the simulation of the fields or the ion channels or whatever is not the same thing as the original, but simply predicts how the original will behave. That's sufficient for intelligence. -- Stathis Papaioannou -- You received this message because you are subscribed to the Google Groups "Everything List" group. To post to this group, send email to everything-list@googlegroups.com. To unsubscribe from this group, send email to everything-list+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/everything-list?hl=en.
