I believe that workable solutions to the problem of improving on unreliable theories is key to AGI at this point. The idea that your data is going to typically give your program precise logical or numerical evaluations as results seems awfully naïve to me. So, assume that the program has developed some weak theories about some situation. How can it improve on those theories? You can see how complicated the problem is when you fully realize (or accept) that if the program is designed to build on other theories, those other theories are also going to be unreliable at first. So let's add that to the presumptions. The program is going to be able to build on previously developed theories about what is going on, and it will try to do that. But now it is working with unreliable theories which are relatively foundational to the current theory under development. So how do you program a computer to deal with this problem? The first thing that come to mind is that the program has to look for results that are more reliable and related to the situation and then these kinds of events may be used as objectives. These objectives may be more specific or transcend the situation under examination but if they can be correlated with other data events that occur within the situation then they might be used as objectives. Furthermore a great deal of attention has to be paid to the structures of the theory. I am not talking about a theory that uses a simple result like a number but of a much more complicated system of theories that somehow act together to bolster the theoretical observations that are correlated with the objective events or explain how some of the events within the situation occur. A causal chain, or more relevant to this problem, various causal relations relative to the situation can provide an example of the structural relationships of the theories that need to be improved on. So when I talk about structural relations I am not talking about the abstractions derived from a foundational computational method like logic or weighted reasoning, but of other insights about the world that are relatively more foundational to the domain of the situation that is being observed and theorized about. Jim Bromer
------------------------------------------- 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
