John,
However, it appears that infinite computation is not feasible, certainly at least not in the short- or medium-term. So, I think what we do is aim at genuine intelligence instead. But now, *given* that our goal is genuine intelligence, I think it is important for many purposes to distinguish between genuine intelligence and infinite computation or Blockhead style "intelligence."
I don't like this kind of distinction between "intelligence" and "genuine intelligence". To me it's like saying that planes don't have "genuine flight" because they don't have some property that birds have. All I care about with regards to intelligence is how well it works. If a machine can cure me of some disease and speed up the development of technology 1000 fold and write computer programs a billion times better than me... and post a few remarkably insightful emails to a few email lists on the side, to me it is intelligent. I really don't care if the machine is a fancy quantum computer or has hamsters running around inside of it. Of course if you want to build a machine with a lot of intelligence (as I define it), then approaching the problem based on the assumption of infinite computation power probably won't get you very far. What you will need to do is to work out how to get as much intelligence as possible out of each unit of computational resource that you have. Once you have done that, you will want to apply as much resource as possible in order to get the maximal intelligence. I've only taken a very cursory look at the AIXI stuff so I didn't want to
talk in any detail about it, but from what I can gather at the moment, that *might* be an example of where this distinction can be relevant. If someone is claiming to be proving some abstract stuff about intelligence but they are really just talking about infinite computation or Blockheadish stuff, then it might be important to keep this distinction in mind and take any claims made about the nature of genuine intelligence with a grain of salt.
For sure. Indeed my recent paper on whether there exists an elegant theory of prediction tries to address that very problem. In short the paper says that if you want to convert something like Solomonoff induction or AIXI into a nice computable system... well you can't. Indeed my own work on building an intelligent machine is taking a neuro science inspired approach with just a few bits that are in some sense "inspired" by AIXI. I think the value of AIXI is that it gives you a relatively simple set of equations with which to mathematically study the properties of an ultra intelligent machine. In contrast something like Novamente can't be expressed in a one line equation. This makes it a much more difficult mathematical object to work with if you want to do theoretical analysis. Shane ----- This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=11983