John McCarthy, for those not in the know, was one of the most brilliant minds in the field of Computer Science and very much likely to go down in history as a foundational thinker of our era.
Absolutely. Beliefs, wants, and intentions are just different types of fuzzy logic. Weak AI—image to text, NPC pathfinding, that kind of thing—is all about imprecise "fuzzy" logic. We call it "artificial intelligence" because we don't have the computing power to get a perfect solution. So we get as close as we can. For example, an expert system for medical diagnosis might "believe" you have a disease with 70% probability. Or, you might assign a robot to go to 10 locations, each with a different level of "want." If it isn't possible to get to all of them in a given time, it can calculate which path satisfies the greatest "wants." The field and geniuses behind AI are so often dismissed because they didn't achieve sentience like they promised. But these guys' research is everywhere. Siri, Wolfram Alpha, financial analysis, game opponents, speech recognition, text recognition, data mining, DNA sequencing. The list is endless.
I was chasing his Elephant and ran into this thread on LtU where the paper was cited. Discussion is brief but substantial if you are interested.
I hadn't heard of Elephant. That is interesting. It would take a bit to convince me it has advantage over traditional formal grammars, or declarative languages. Algol derivatives and, you know, every other structured programming language. "Refer to the past directly" sound suspiciously like versioning. In fact, most of the language sounds like a bunch of high-level concepts and abstractions smooshed together. The last time someone tried that, we got C++.Algolic programs refer to the past via variables, arrays and other data structures.
Its not merely about syntax. Of course John is RIP so I am guessing. I was chasing elephant since I read an interview of his yesterday (infoQ) and he was talking about the concept, and it sounded very much like what I am working on now. The underlying system is a (hopefully -- shopping for consensus algos now) linearly scalable database. (Cap:AP semantics.) The persistence layer is content addressable, but the system is Temporal (Yes, MVCC). It has a 'theory of minds and collective knowledge' that informs the design. Key point/insight is realizing that f (t, k) -> v can support a lambda calculus (LISP, though Prolog as query language would work as well). Bonus: Since every computation is on past -- isn't that how your brain works, too? :) -- IO does not require monadic gymnastics. And in honor of JMC, have decided to call the (maybe one day happening) language on top of it -- think a distributed LISP Machine -- FEAL-T. Feal means Elephant in Persian and T is for Temporal and Fealty is a characteristic of Elephants ..