Matt Mahoney <[EMAIL PROTECTED]> wrote: As you might know, my interest is data compression. Its relation to AGI is explained in http://cs.fit.edu/~mmahoney/compression/rationale.html To summarize, you need to solve AI to compress text well. I don't pretend to know how to do that. But I do claim that it is a better way to measure intelligence than the Turing test.
I think AGI is a poorly defined goal. You can't ever say if it has been achieved or not. Useful intelligences will not be like the human brain with all its weaknesses, like emotions, biases, slow response times, spelling and arithmetic errors, and limited short term memory that seem necessary to pass the Turing test. If your goal is to use AGI to replace all human labor, then it is only necessary to make it intelligent in areas related to its job. Google far surpasses humans in its ability to answer simple questions, but you would not call it AGI. Google doesnt ditectly answer anything, just provides helpful pointers, but does that fairly well, given a good intelligence controlling the search terms. I would really like to see a coordinated list of goals come out of this list. I have seen a lot of talk and concern about the Singularity and friendliness and super-AI, which is frustrating to me, cause we need to concentrate on creating a basic or working AI first before it can and we can worry about a super AI. A simple list, or set of goals for an AGI to accomplish reasonably I would find very useful, and something to work for. Most things I have seen consist of either specialized expert system types, or "General" intelligence where it is defined as "Do Everything" that humans can and more and better. We need to have a well defined set of possible goals in between where we can work towards those as opposed to being bogged down tryign to create a super being. To that end, I propose an AGI should be able to act as a normal human common, 'stupid' person. I dont worry in the beginning about it coming up with any kind of unique wonderful ideas. One test of this that I am intersted in, is putting the AI in common environments and situation from fiction novels and allowing them to act, and for you to interact with them, for them to learn there and accomplish simple goals. more later... James Ratcliff _______________________________________ James Ratcliff - http://falazar.com Looking for something... --------------------------------- Ahhh...imagining that irresistible "new car" smell? Check outnew cars at Yahoo! Autos. ----- 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/?member_id=231415&user_secret=fabd7936
