> > The BIG question is: What good are a lot of 95% "facts". You can't rely on > ANYTHING, and as soon as you start putting those "facts" together in > combination, the accuracy falls WAY below 95%. The more "facts" that are > strung together, the less accurate the results. For example, the likelihood > of just 10 x 95% "facts" all being correct is only 60%., and the likelihood > of getting 20 correct is only 36%. Hence, any "AGI" applying limitless > computing capability to make sense of Wikipedia is only going to generate a > lot of gibberish, possibly including some jewels, but without any capacity > to separate the jewels from the broken glass without access to the real > world.
This only makes sense if the probabilities of the list of "facts" cannot be leveraged for consistency-based error correction. If I can take two 95% likely facts and use them to identify a third which is inconsistent with them, I can often correct that third one based on how it conflicts with the first two. It doesn't make for guaranteed 100% accuracy, but it can significantly improve the error rate. It's a similar principle to boosting in machine learning. http://en.wikipedia.org/wiki/Boosting_(machine_learning) This is why I think it's important to integrate knowledge. On Sat, Mar 30, 2013 at 5:57 PM, Steve Richfield <[email protected]>wrote: > Mike, Jim, andBoris, > > You are halfway there. Here is the other half: > > When analyzing text, no man or machine can be 100.000% confident that they > precisely understand what the speaker INTENDED to communicate. A couple of > percent are flat-out erroneous. Maybe another 3% are beyond any presently > conceivable practical analysis. Even shallow analysis can help identify > which statements will have higher probabilities of analysis error, e.g. > more words -> more errors. Few statements would could be analyzed with a > confidence >99%. I suspect that 95% would be around the upper limit of > conceivably achievable by practical means. Just picking out certain items > of interest and ignoring everything else, DrEliza gets it right ~90% of the > time, but restatement of facts improves this number. > > The BIG question is: What good are a lot of 95% "facts". You can't rely on > ANYTHING, and as soon as you start putting those "facts" together in > combination, the accuracy falls WAY below 95%. The more "facts" that are > strung together, the less accurate the results. For example, the likelihood > of just 10 x 95% "facts" all being correct is only 60%., and the likelihood > of getting 20 correct is only 36%. Hence, any "AGI" applying limitless > computing capability to make sense of Wikipedia is only going to generate a > lot of gibberish, possibly including some jewels, but without any capacity > to separate the jewels from the broken glass without access to the real > world. > > Boris was making the same point that I tried to make - that you must > design to an APPLICATION, or at least design to serve certain applications. > I don't see how any super duper "AGI" language understanding program is > ever going to do anything useful at all, in the face of ordinary > unavoidable human and language-design error rates. Nonetheless, lots of > apparent "nuts" have worked for decades on just such programs - with no > apparent success, and from what I can see, no realistic hope of ever > succeeding, even if given limitless resources. If there is ANY realistic > basis for this hope, I would sure like to hear it. > > Steve > ============== > On Sat, Mar 30, 2013 at 11:47 AM, Mike Tintner > <[email protected]>wrote: > >> *PS. To summarise: “only real world agents can use language” - >> “only agents engaging with the real world and its objects, can understand, >> actuate and assess language’s statements about the real world and its >> objects” - and **language, in the final analysis, consists of nothing but >> statements about the real world.*** >> ** >> *Even language statements about fantasy or alternative worlds are ipso >> facto statements about the real world – about objects that are different >> from - and therefore implicitly describe - those of the real world.* >> *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> >> <https://www.listbox.com/member/archive/rss/303/10443978-6f4c28ac> | >> Modify <https://www.listbox.com/member/?&;> Your Subscription >> <http://www.listbox.com> >> > > > > -- > Full employment can be had with the stoke of a pen. Simply institute a six > hour workday. That will easily create enough new jobs to bring back full > employment. > > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/23050605-2da819ff> | > Modify<https://www.listbox.com/member/?&;>Your Subscription > <http://www.listbox.com> > ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657 Powered by Listbox: http://www.listbox.com
