WITH REGARD TO THE FIRST QUESTION BELOW: I did not in any way suggest that I was going to cripple the AGI I was talking about, nor did I suggest their skills would be limited to those comparable to humans. "Common sense" as used in AI usually refers to a very important form of reasoning, actually much more important in most contexts that what many have traditionally labeled as "logical" or "rational" thinking. Common sense is the ability to reason from large amounts of knowledge, often knowledge involving uncertainty and only approximate matches, in a context appropriate way. I consider common sense to be both logical and rational, just that it often involves performing logic and rational thought in complex, uncertain, less clearly defined situations -- where many traditionally defined systems for logical or rational thought totally fail. And, in many cases, much, if not most of its functionality, is carried out, at least in humans, in the subconscious.
To a large extent creating a system that can automatically learn and compute common sense from a complex world is one of the key tasks of AGI. There is no reason a powerful AGI could not have vastly superhuman common sense. (Although such common sense would not be "common" among most people.) WITH REGARD TO THE SECOND QUESTION BELOW (which seems to understand that "common sense" can be superhuman as I have stated above): it depends on the definition and test for "mentally health" when applied to a machine, which is a wide open question. Being something of a humanist (although one who believes in the ultimate need for transhumanism) I would initially define machine health largely in terms of usefulness, capability, reliability and safety, all from a human viewpoint. But over time this viewpoint would probably change. There is no reason why rationality and common sense should conflict, since as stated above common sense is rational. Ed Porter -----Original Message----- From: Mike Dougherty [mailto:[EMAIL PROTECTED] Sent: Wednesday, November 07, 2007 1:43 PM To: [email protected] Subject: Re: [agi] Human vs human-level Intelligence On 11/7/07, Edward W. Porter <[EMAIL PROTECTED]> wrote: > There is no reason why properly designed AGIs with world knowledge and > the power to compute from it would have any less common sense than > humans. Ok. Are you also going to sufficiently cripple your AGI's ability to think rationally that they are completely comparable in skills as a human? With super-human skill at "common sense" and equally superhuman rationality, will this AGI be considered mentally healthy if the observing psychologist is not augmented to extra-super-human reasoning? ----- 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/?&; ----- 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=8660244&id_secret=62564912-55c6b7
