Aside from Ben's reply (which suggests talking to Nil about his work on temporal reasoning in PLN) I can add in my 2 cents, which is: nothing that I've worked on deals directly with time in any way :-)
> It's hard to tell though what is speculative theory, what is implemented I can be very precise about this, if you ask the right questions. Very briefly: * The atomspace is very mature, debugged, performance-tuned and stable. This includes the query engine, the 120+ Atoms that constitute Atomese, the notion of "Values", the distributed database backend, and the "matrix" layer (it can take arbitrary slices of the AtomSpace contents and expose them as a sparse matrix (of rows and columns), thus allowing conventional probability theory to be applied (e.g. conditional probability of row given column, etc.) The Atomspace has been used/applied in various ways in at least a dozen different projects (that's how it got mature - trial by fire). * The Unified Rule Engine (URE) is mature, but has only been used in one project: PLN. It was supposed to provide a generic way of running rules, but no one aside from PLN ever used it for that. (one of half-a-dozen examples from a long time ago was to put the NARS rules on top of URE) * PLN is a specific collection of rules. Some of these are fully worked out and mature; some are experimental, some are in active development. Nil can say more. * I'm very actively developing code for generalized learning (currently, language, in general, vision and sound; yes there's a unified theory that covers all this and more.) * Everything else is immature, incomplete, abandoned and/or bit-rotted. This includes over a dozen different subsystems, one of which is a "spacetime server" that can record things/events/generic-stuff in space and time. Some of this stuff should be brought back from the dead, some should be allowed to rest in peace. I've one comment about time and temporal theory, and I suspect you'll hate it. I strongly believe that an AGI system must learn about time, (and how to reason about events in time) instead of having it hard-coded into it. Likewise, it must learn "common sense", and only after that, can it learn about rationality and reasoning. Thus, I spend all my effort on learning; I expect it to (eventually) learn common sense and how to reason about time. This is why I (personally) don't work on any explicit theories of time. This differs from older (dare I say "conventional"?) architectures, where all this stuff (space, time, reasoning, logic) is hard-coded in at a base layer. -- Linas On Wed, Feb 2, 2022 at 3:50 PM Mike Archbold <[email protected]> wrote: > I'm doing some research for the group I organize, the Northwest AGI Forum. > I'm trying to get a feel for how advanced time theory is in architectures. > Time obviously is important. > > I'm familiar with OpenCog in general terms, and over a the years I have > studied it, although I don't know the details. I was browsing through the > wiki. I see references to modal logic, PLN, the atomspace, etc. It's hard > to tell though what is speculative theory, what is implemented, what is > planned, long term envisioning etc. I know Ben et al has done a lot of work. > > So I was wondering if someone could please give a bit of an overview as > far as the general state of time representation in OpenCog, or perhaps > point me to the relevant literature. > > I really appreciate any help! I want to summarize time theory in OpenCog, > NARS, and SOAR. > > Thanks Mike Archbold > > > > -- > You received this message because you are subscribed to the Google Groups > "opencog" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To view this discussion on the web visit > https://groups.google.com/d/msgid/opencog/abe64f28-da56-4ec4-84d5-8c51d3eab78dn%40googlegroups.com > <https://groups.google.com/d/msgid/opencog/abe64f28-da56-4ec4-84d5-8c51d3eab78dn%40googlegroups.com?utm_medium=email&utm_source=footer> > . > -- Patrick: Are they laughing at us? Sponge Bob: No, Patrick, they are laughing next to us. -- You received this message because you are subscribed to the Google Groups "opencog" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/CAHrUA37sWFaCMssDKHzfLnYuv%2B4sM%3DR_F2HfR-X%2B4uH9ou48BA%40mail.gmail.com.
