Can Watson be used as the basis for AGI? The problem is that certain word-phrases (or other kinds of observation events) have functional-conceptual implications. (I believe that almost all concepts have broad functional implications). It should be fairly easy to implement these functional effects if you keep them down to a minimum or you develop some kind of abstract basis for these kinds of phrases which can be kept to a minimum. Watson, with its Deep NLP might be able to keep track of these kinds of subject-relative words (that have strong functional implications). However, Watson excelled just because it was able to find the right answer to certain kinds of questions. Is IBM going to want Watson to develop mistakes that it then has to learn to correct over long periods of conversation? The fact that they are willing to present humorous videos of Watson failing to be perfect does show that the company is not uptight about this. The diversity of human beings seems to partly explained by the fact that we can develop our own opinions (and oops-pinions) and then just run with them to see how useful they can be. With massive server memory ready to use this should not be a problem for a research team.
But there is certain knowledge that is critical for elementary success. Simple recognition is one step. But the next step requires that AGI goes beyond expected discrimination abilities and still stay on track in some ways. The problem with AI has that it has totally lacked sufficient traction even to scale up to be a little human like. The necessity to get certain kinds of knowledge right in order to make real progress on a more human-like AI might turn out to depend on a well structured set of abstractions that have not been developed yet. I am a little skeptical of this. I think these abstractions would have to form the basis of what I am calling Conceptual Integration. But, as I have said, Conceptual Integration will have to be - at the very least - governed by concepts, and - more reasonably - they are concepts themselves. Jim Bromer On Thu, Jan 14, 2016 at 6:52 PM, Mike Archbold <[email protected]> wrote: > Jim: Even though Watson-Jeopardy did not use Neural Networks or something > that was intuitively similar to them, I believe it was an example of > deep learning. But the question that many of us are more interested in > is was it an example of Narrow AI? My first response is that it is not > because it can be applied to such a wide range of problems (even out > of the box-or out of the virtual box). So then, why isn't it AGI? Why > can't it think outside the box? Why does it not demonstrate the traits > of what I call semi-strong AI? This question bothered me but I think I > finally have figured it out. > > Channelling Mike Tintner: "It's not AGI because it isn't creative. > It's just answering questions. To be human-like AGI IT HAS TO BE > CREATIVE. Meaning given some novel situation, it comes up with a > solution that was not preprogrammed." > > The answer to that criticism is yes, Watson wouldn't be creative in > that right, but maybe they built in or will build in some evolutionary > functions to approximate "creativity." > > Anyway, I like Ben's list of required attributes to be AGI, which I > don't have at hand. I am pretty sure creativity is on the list. > > A lot of this discussion is around if some system is or is not AGI, is > it narrow-AI, or the like. I think we have to allow that "less > narrow" is a possibility. I know AGI is qualitatively different, but > still some approaches seem to push beyond narrow AI. Implementing > less narrow AI is obviously within the realm of possibility. Watson > sure looks less narrow!!! But, then again, we are left with > reclassifying "less narrow AI" as narrow AI, moving the goal posts.... > this has been going on for a long time. > > But, I think a crucial question is if you could implement a great less > narrow AI, could that serve as a basis and nucleus for authentic AGI > later on down the line? Or do you have to throw out your less narrow > AI nucleus? > > Mike A > > On 1/14/16, Jim Bromer <[email protected]> wrote: >> I realized that Deep NLP that Watson probably referred to Deep Search >> NLP because there was something about examining text to find relations >> between words in specialized cases as well as the most general >> relationships that are found by taking only the most frequent >> relations. (My paraphrasing is terrible but that is the idea that I >> came across In one of the IBM texts about Watson.) However, there is >> more to it than that of course. And the discovery of specialized >> relations is a kind of learning. >> >> >> ------------------------------------------- >> AGI >> Archives: https://www.listbox.com/member/archive/303/=now >> RSS Feed: https://www.listbox.com/member/archive/rss/303/11943661-d9279dae >> Modify Your Subscription: >> https://www.listbox.com/member/?&; >> Powered by Listbox: http://www.listbox.com >> > > > ------------------------------------------- > AGI > Archives: https://www.listbox.com/member/archive/303/=now > RSS Feed: https://www.listbox.com/member/archive/rss/303/24379807-653794b5 > Modify Your Subscription: https://www.listbox.com/member/?&; > Powered by Listbox: 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
