The limited goal I am pursuing is not to answer many questions but to be able to show that it is capable of understanding (or acting like it is capable of understanding) something. To do this it not only has to be able to search for a piece of information that matches some key words, some grouping of key words or some learned categorization based on the inclusion and/or sequencing of key words, but it also has to be able to interpret something from the (limited) domain of the distributed data. The key here, I believe, is that some data (like words) can be interpreted as instructions (for the program to subsequently use) and these abilities can lead to some primitive judgement. So, my opinion is that the early efforts at limited domain AI did not work because the programmers did not have the insight (or the computing power) to be able to write programs that could learn to 'read' sentences and interpret them as actions for the program itself to take. Of course this ability would lead to more complexities but given the fact that I completely accept that my program would - at best even if it did work - be seriously limited I should have no problem limiting that particular kind of complexity.
If it is indeed possible for me to write a crude version of this program it might not be the first program that showed some ability to interpret a sentence that was not based on an severe uniqueness constraint so that it could use that interpretation as guidance toward an action but it would be among the first that was carefully crafted to do so. There are quite a few characteristics that I believe are necessary for 'higher' general intelligence (even if limited higher intelligence.) The ability to learn to interpret a sentence (or a concept) as guidance for its own actions is only one of them. It has to be able to work with non-unique symbols and it has to be able to learn new strategies by developing new types so that it can create virtual sub-programs based on its learning. And it has to be able to utilize judgement which has been learned, which it has read (or 'read' from its observations), and which it can derive from experience. Jim Bromer On Tue, Dec 11, 2012 at 8:32 PM, Matt Mahoney <[email protected]>wrote: > On Tue, Dec 11, 2012 at 3:11 PM, Jim Bromer <[email protected]> wrote: > > > > I was just asking Google and Bing questions and I was surprised at how > well they did. > > Yes, it is amazing to see how far search engines have come, and how > close they are to AI. One thing that they have and you don't is enough > computing power to keep a copy of the internet in RAM. > > -- > -- Matt Mahoney, [email protected] > > > ------------------------------------------- > AGI > Archives: https://www.listbox.com/member/archive/303/=now > RSS Feed: https://www.listbox.com/member/archive/rss/303/10561250-470149cf > 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
