Brief: In a directory on a Debian system there lived a hobbit...
Jump To: Parent Description

  • So, what is seed ai? Although seed ai suffers from a transitive closure violation unless one assumes a friendly environment, it is worth developing such a system to speed things up. In other words, most programming that we are doing is declarative, there is no real inferencing done by the software. Now, we want the system to do a lot of its own reasoning. We don't need it to think so intelligently that it can reason with absolute precision about certain problems - in fact, the quid pro quo error of mathematics is that very often something that is rejected as not being a solution may have 100 % of the solution. When this is desirable is another constraint. So, the same type of thinking is necessary here. What we want out of it is introspection and a novel amount of reasoning, but mainly, independence, gallant behaviour etc. We would like it to solve its own problems, to get bored and switch to others. Mostly, to seek out answers to problems and fixes. The large mechanisms for perception are readily available through our comprehensive perl bindings. But the ability to think and act, not as an adult (lifting A.I. by its brooches so to speak), nor precisely emulating biological cognitive structure, nor a tediously exact code implementation. No, it should be a complex little cognitive architecture, with a visual buffer, communication, etc. No draconian formalisms here. No exactly replication of behaviour. But a seeker of knowledge. Definitely like PIC in its complexity but not as exact... It can get stuff done because, although in exact, it is not open ended like conceptnet, but self contained, constituting of most flavors of cognitive systems theory without much to go. Kind of a hobbit.