Hi Ivan, I liked the name "Fossils" so I went ahead with that. There are now two brand-new wiki pages:
* https://wiki.opencog.org/w/OpenCog_Fossils * https://wiki.opencog.org/w/OpenCog_Incubator For each of the fossils, I attempted to write a post-mortem: what the idea was, what went wrong, why it didn't quite work out as hoped. I think post mortems are good: they give you time to think over what happened, to draw some lessons. A time to ponder how things could be done differently. These are wiki pages! Those who don't agree, or have something to add, or can deduce a different set of lessons -- well, this is the time and the place for that. There's one political statement in the middle of this: a proposal to layer MeTTa on top of the AtomSpace. From what I can tell, this would be easy to do, and, from what I've seen, you'd get a huge performance boost and instant scalability, since the AtomSpace can deal with far larger datasets than what Hyperon can currently handle. I know that this is an unpalatable pill, and will be rejected out of hand, ... but perhaps it's time to think a little more seriously about things. FWIW, the proof-of-concept is at https://github.com/opencog/atomspace-metta -- Linas On Fri, Jun 30, 2023 at 7:29 PM Ivan V. <[email protected]> wrote: > Hello dear OC community, > > May I drop just an observation as an independent eye? How about something > constructive like (based on Linas' suggestions): > > 1. OpenCog Genesis > 2. OpenCog Hyperon > > and additionally: > > - OpenCog Fossils > - OpenCog Incubator > > --- > > Use for Fossils is obvious, and relates to Linas' excellent idea. A lot of > people tied up a valuable time in it, and holding that work in an archive > like this would be a great. Who knows, maybe someone has a clue how to > revive portions of those extensions. > > Genesis and Hyperon would be the first two incompatible versions in a row: > > - Genesis would be the current version (filtered at someone's wake > eye); it already works, it already has a user base, and it is given enough > attention not to just let it go like that; just package it right, and I'm > sure it will have enough users. Something should be made out of it, and I > like Linas' propositions. > - Hyperon would be (as a product of a lot of experimentation - not to > be disregarded) an evolutionary step from Genesis; obviously, a > considerable amount of effort is invested in it; it is the second, but > surely not the last step in this mainstream timeline. > - And the following versions... boy, I can't wait to see what the > future brings... the future is always surprisingly ingenious... > > > As a potential independent contributor, I'm particularly interested in > Incubator as a place from which all the mainstream versions would be > updated and refreshed by extensions worth of consideration. I like the OC > openness, but I find the current strategy too chaotic to hold everything > together in a right place. Incubator could solve that problem in my humble > opinion. I wish there could be a clearer connection from Incubator to > Genesis and Hyperon in terms of GitHub, but what we have, we have, and I'm > grateful for it. > > @Linas, do you have any thoughts on these suggestions or am I being too > insolent? > > love, > Ivan > > Dana nedjelja, 18. lipnja 2023. u 04:20:50 UTC+2 korisnik Linas Vepstas > napisao je: > >> Naming and branding: >> >> So here's a proposal. Four things. Four names. Four brands: >> >> * OpenCog Classic >> * OpenCog AtomSpace >> * OpenCog Hyperon >> * OpenCog Incubator >> >> "OpenCog Classic" consists of everything abandoned and obsolete; This >> includes PLN, URE, Attention, SpaceTime, Ghost, Relex, R2L, ROS, >> Eva/Sophia, MOSES (but not as-moses, see below). >> >> "OpenCog AtomSpace" includes: >> -- base AtomSpace >> -- CogServer and atomspace-cog (for networking, json, websockets) >> -- atomspace-rocks (disk I/O subsystem) >> -- Proxy Nodes (for data routing, replaces attention bank) >> -- Link Grammar >> >> OpenCog Hyperon >> -- the stuff you're working on >> >> OpenCog Incubator >> -- atomspace-prolog proof-of-concept >> -- chemistry proof-of-concept >> -- agi-bio >> -- visualization >> -- as-moses >> -- vision proof-of-concept >> -- hyperon-on-top-of-atomspace proof of concept. >> -- etc. >> >> The above is the branding and website design that I would like to see. It >> makes it clear what's what, what goes where, what's alive, what's dead. >> what's stable, what's new. what's broken, what's experimental. Does this >> work for you? >> >> Two extra remarks: >> >> The hyperon-on-top-of-atomspace proof of concept. Not sure if you didn't >> see the announcement, or you saw it and just hated it. I think putting >> Hyperon on top of the AtomSpace is not too hard, and I think it might work >> quite well, given that the code base is debugged and performance tuned. I >> understand that the current devel crew does not care for such a thing. But >> it is a viable path, and a kind of insurance policy. >> >> Link Grammar. I've been working to integrate Link Grammar more closely >> with the base AtomSpace. It turns out that Link Grammar is a generic >> structure parser. Or rather, is close to one, although some fair amount of >> bending, welding and hammering is still required. I think it can be used to >> process visual/video data, audio data, sensory data in general. Part of the >> big deal is that link-grammar is now really really fast: Amir Plivatsy has >> performance-tuned the heck out of it, it runs about 100x faster than it did >> a decade ago. And since CPU's have also gotten faster, ... wow. So that is >> the dataflow, or structure-flow idea. Converting any kind of streaming data >> into a "streaming semantic graph" in real time. >> >> AS-MOSES: As a whole, it is problematic. There are a few pieces/parts I'd >> like to rip out and breathe new life into. In particular, generic >> "knob-turning" and deme management would be excellent stand-alone plugins >> for the AtomSpace. I need knob-turning in order to train on video and audio >> feeds. >> >> -- Linas >> >> >> On Thu, May 18, 2023 at 6:18 PM Ben Goertzel <[email protected]> wrote: >> >>> I think Linas may not care for the appellation "Classic" for the >>> version of OpenCog he's currently developing and. using .. I don't >>> really love it either but I just came up with it casually and have >>> kept using it.. it was kind of a bad joke on "coca Cola Classic" >>> really... >>> >>> Linas, how would you suggest we refer to the original, non-Hyperon >>> version of OpenCog on the opencog.org site... given that we want to >>> clearly point folks who come to the site both to a page giving >>> resources on Hyperon, and to resources on the other original/classic >>> OC version (e.g. the OC wiki, the Github repo, etc.) ... >>> >>> thanks >>> ben >>> >>> On Thu, May 18, 2023 at 12:46 PM Haley Lowy >>> <[email protected]> wrote: >>> > >>> > Hello Linas! >>> > >>> > We are planning to have our web person start with a refresh of the >>> look and feel of the OpenCog front page - drawing largely from the content >>> that is on the foundation site currently about the theoretical >>> underpinnings of OpenCog and its vision and goals. We’ll then have a >>> section that features the two versions - “Classic” (which is how we’ve been >>> referring to the original), and “Hyperon”. We will link out to the wiki, >>> the discord, and the google group for those who want to get involved. >>> > >>> > We will then, as Ben says, add a subdomain for Hyperon - we currently >>> don’t have a good place to collect and highlight the features and progress >>> of this project, so this page will allow us to do that. >>> > >>> > Mario Guzman will be doing the website design and development, and we >>> are currently building wireframes for it, which we should be able to share >>> early next week. We are hoping to have at least a fresh landing page and >>> some initial resources up before AGI-23. >>> > >>> > What do you have in mind, and what are your goals for a new site? >>> > >>> > I look forward to getting to coordinate with you on this, as your >>> passion and enthusiasm for OpenCog has been so helpful to devs and >>> researchers who want to get involved in using this system. Thank you, >>> > >>> > Haley >>> > >>> > >>> > > On May 16, 2023, at 5:40 AM, Ben Goertzel <[email protected]> >>> wrote: >>> > > >>> > > Hey! >>> > > >>> > > It seems that two parties (Haley / SNet team) and Linas are >>> > > concurrently thinking about updating the OpenCog.org site in the near >>> > > term... >>> > > >>> > > Def. this update makes sense, the site is more than a bit stale... >>> but >>> > > let's make sure we're coordinated here... >>> > > >>> > > Linas: Our plan has been to make a subdomain hyperon.opencog.org and >>> > > use it as a main info page on Hyperon, and then insert a link on this >>> > > on the (revised) front page. I suppose this would be compatible w/ >>> > > whatever changes you're thinking of making? >>> > > >>> > > Anyway I don't want to be in the middle here communication-wise so w/ >>> > > this email I am aiming to put Haley and Linas in direct communication >>> > > on this! >>> > > >>> > > thx >>> > > ben >>> > > >>> > > >>> > > -- >>> > > Ben Goertzel, PhD >>> > > [email protected] >>> > > >>> > > "My humanity is a constant self-overcoming" -- Friedrich Nietzsche >>> > >>> >>> >>> -- >>> Ben Goertzel, PhD >>> [email protected] >>> >>> "My humanity is a constant self-overcoming" -- Friedrich Nietzsche >>> >> >> >> -- >> 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/f4592451-f1ff-4e73-b608-7db356e83f4dn%40googlegroups.com > <https://groups.google.com/d/msgid/opencog/f4592451-f1ff-4e73-b608-7db356e83f4dn%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/CAHrUA368Z7u8kJLuktSGNLj-3mXBnkLh4xBkc5v65S5pEr_GQg%40mail.gmail.com.
