Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!swrinde!zaphod.mps.ohio-state.edu!wuarchive!udel!mmdf From: doug@ctc.contel.com (Doug Whitehead x4149) Newsgroups: comp.sys.amiga Subject: Re: Amiga and AI Message-ID: <13174@baldrick.udel.EDU> Date: 7 Mar 90 16:47:18 GMT Sender: mmdf@udel.EDU Lines: 45 I work for Contel, a telecommunications co. and am a PHD student at VPI. I am using my Amiga for development of a new language for hypothetical reasoning. (related to truth-maintenance) I have previously implemented a version in Prolog on a Vax, but the overhead associated with an interpreted prolog was rediculuous. The new version is in C. The sort of reasoning this new language will capture is along the lines of the following example: A robot is in room1, a box is in room1, the robot is holding the box. If I were to introduce a hypothesis (something believed above all else) that the robot is actually in room2, then under "normal" circumstances, the computer should suggest two possible worlds with this hypothesis: A) The box is also in room2, or b) The robot is no longer holding the box Of course, if you wish to relax constraints and allow the computer to be very skeptical, other explanations are possible, such as: Room1 = Room2 Robots can be in two places at one time Robots can hold things in other rooms The normal definition of "=" is incorrect etc. Obviously I neglected to show you the rules that connect the facts of any world. rules like: If X holds Y and X is in W, then Y is in W As you can see, this language is suited for intelligent databases (or knowledge bases; intelligent in the sense that the infobase is consistent with itself) and qualitative physics (If I lift a sauser, the teacup, and the tea should normally move with it, etc.).` Anyway, as for your original question, there is an excellent prolog ported to the amiga, SBProlog (Stoney Brook Prolog) is on one of the fish disks around 143 or so..., (Be careful TinyProlog on one of the surrounding disks sucks...) SBProlog is a full interpreter/compiler, a rare treat even for commercial products. Personally, Prolog has most of what one wants in a shell (unless you need to dazzle a client/boss, in which case graphics are important, not expert system content). Don't be bamboozled by "certainty factor" or probability rhetoric, they are generally considered unuseful in expert systems, (at least by those who have built an expert system before) If you have any specific questions, I'll try an field'em 'Haid