I don't believe that major advancements in AGI are feasible right now. However, major advancements in computer technology are occurring all around us so the sheer power of the hardware means that some things that had been impossible before are now possible.
One of the most amazing things I have seen during the past few years are videos of helicopters that have learned to fly upside down and do all sorts of tricks like flying through small openings. These remind me of the amazing aeronautical feats that birds can achieve (like high-speed flight through small areas between branches.) I believe these helicopter tricks were learned using neural networks. And I believe that the reason that neural networks were successful for these and other projects like learning to recognize drawn characters was because the method was so simple. They were used to achieve some things that more elaborate methods were unable to achieve because they did not run into the complexity barriers that more complicated methods can create. The reason that neural networks have not been used to successfully learn more interesting things is because they are too simple. So I believe, based on my limited experiences actually trying to write algorithms that were supposed to do thorough pre-analyses of a situation, that the contemporary methods that are more likely to produce some limited successes (that might look like benchmarks of progress) will be methods that are a little more sophisticated than neural networks but not too much more sophisticated. Most people here would think of probability nets. I have so many problems with the conventional thinking about probability nets that I don't agree only because the adherents don't seem to acknowledge the logical dilemmas of the approach. So I would conclude, based on these reasons, that it is more likely that logic nets will produce some limited breakthroughs in the near future. Strangely, logic nets in AGI have not been discussed widely so that suggests that the field is wide open. So I am saying that the field of logic nets in AGI is both unlimited and yet to be truly defined. There were some attempts in the old days, like constraint-based reasoning (I have to look it up to remember what that was) and semantic nets but these were only two possible variations. The term logic net might be misleading. Perhaps I should name the method relational nets. Yes, that sounds much fresher and it gives you a sense of how much freedom it might provide programmers because it is not just based on logic but the application of methods to networks. (Ok, I found a definition of case-based reasoning so now I should remember what that means for a few days.) Jim Bromer On Thu, Aug 30, 2012 at 4:53 PM, Anastasios Tsiolakidis <[email protected]> wrote: > Well folks, last time I took a mid-level look at your designs, asking > for your non-negotiables, the results were unimpressive. As our > ability to do benchmarking, incremental progress, milestoning etc > remains limited, I would like to ask again for your eagle's eye view: > > 1) which breakthrough/milestone you think is/are needed on your way to AGI > 2) which module(s)/subsystem you are certain will make it all the way > to AGI 1.0, so which current theory/tech is essential to your vision > > I would like to cop out of 2) with "logic", i just don't see AGI 1.0 > arriving without some explicit Aristotle or Prolog inside it, it is > the computational shortcut to end all shortcuts. Less powerful but > more general, some probabilistic system will be there too, possibly > multiple ones including fuzzy logic. I'd also wager a bayesian > element is de rigeur. I am not trying at all to be exhaustive here, > just giving actual examples. > > 1) is the more critical and open ended, I sincerely think that the > "correct" agent-like distributed architecture is disruptive enough to > be called a breakthrough towards AGI, but I am not sure at all I know, > or anyone knows, how to avoid the curse of dimensionality, the curse > of productivity, the curse of contradiction and the curse of > non-terminability. > > AGI could be cursed, folks! > > AT > > > ------------------------------------------- > AGI > Archives: https://www.listbox.com/member/archive/303/=now > RSS Feed: https://www.listbox.com/member/archive/rss/303/10561250-164650b2 > 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-c97d2393 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968 Powered by Listbox: http://www.listbox.com
