Yes, my current impression agrees with Linas's.... This does look like the same stuff we reviewed privately a year ago.
It's sort of like someone made a UML diagram for the human mind, then created a software object for each part of the mind, and then filled in each box with some hard-coded stuff that sorta seemed to fit the bill in some sense.... When Minsky was trying this in the 70s at least it had the merit of originality. But by now we should know that doesn't really lead anywhere. The UML diagram is OK but you need to fill in each box with extremely flexible representation systems and powerful learning algorithms, and then make all these representation systems and learning algorithms interoperate synergetically... In other words this really looks to me like straight-out GOFAI using modern software tools... -- Ben G On Sun, Jul 16, 2017 at 4:51 AM, Linas Vepstas <[email protected]> wrote: > And one more thing, then: > > https://github.com/watson-intu/self/tree/develop/src/blackboard > > is a "blackboard" -- see > https://en.wikipedia.org/wiki/Tuple_space > https://en.wikipedia.org/wiki/Blackboard_(design_pattern) > https://en.wikipedia.org/wiki/Blackboard_system > > Note how everything in the black-board is hard-coded: person, health, > entity, image, gesture, etc. > > In opencog, we have a blackboard too, its called "the atomspace". It can > store arbitrary data in a persistent fashion. > http://wiki.opencog.org/w/AtomSpace > > The atomspace stores very low-level objects, that roughly resemble "assembly > code": the kind of stuff a compiler might act on. There's actually a lot of > compiler-like stuff in the atomspace. There's a reason for that: we want > high-level algorithms to be able to examine and alter patterns, and to learn > brand-new patterns. Thus, we do NOT want to fix the representation of > "person, health, entity, image, gesture,etc" into a frozen, unchangeable C++ > object that only human programmers know how to modify. We want opencog > systems to learn, on their own, what people, health and gestures might be, > and alter these definitions over time. (De facto, we do hard-code a lot of > stuff like that, but the long-term goal/design is not to do that) > > Also the atomspace can be: > -- saved to disk > -- can be searched and queried > -- can hold rules, and comes with a rule engine > > The watson-intu/self system can't do any of these things, as far as I can > tell. Let me know if I'm wrong: I love to rip off ideas: "imitation is the > sincerest form of flattery" if there's something worth imitating, let me > know. I am humble enough to just-plain "ape" if I can't imitate. > > --linas > > On Sat, Jul 15, 2017 at 3:14 PM, Linas Vepstas <[email protected]> > wrote: >> >> Maybe a year ago, I got a brief pitch of some Watson robot code, to which >> my reaction was "OMG they are reinventing ROS from scratch, badly". My >> super-quick skim of this code suggests that its the very same code. I might >> be very wrong, but I looked at this: >> >> https://github.com/watson-intu/self/tree/develop/src/sensors >> which offers camera, microphone, sonar ... about 4-5 sensors. vs. ROS >> offers hundreds or more? >> >> https://github.com/watson-intu/self/tree/develop/src/models >> which offers definitions for a vertex, edge and graph. WTF. This seems >> like an undergrad homework problem. What am I missing? >> >> The planning code here: >> https://github.com/watson-intu/self/tree/develop/src/planning >> is only slightly more extensive, but seems to be a freshman attempt, >> easily outclassed by 30-year-old software like >> https://en.wikipedia.org/wiki/CLIPS >> >> A survey of modern approaches is here: >> >> https://en.wikipedia.org/wiki/Logic_programming >> https://en.wikipedia.org/wiki/Motion_planning >> https://en.wikipedia.org/wiki/Constraint_satisfaction_problem >> >> I randomly picked through the other directories, and ... was not >> impressed. My general impression is that its like something one might have >> seen in the 1980's-1990's, in the dark of AI winter. SGI had whizzier stuff >> back then, whizzier stuff could be found at SIGGRAPH. Am I missing the >> forest for the trees? Is there anything here? >> >> --linas >> >> On Sat, Jul 15, 2017 at 10:24 AM, Ben Goertzel <[email protected]> wrote: >>> >>> Having code "for" all those things doesn't mean much really does it .. >>> it's easy to create a bunch of OO classes with fancy >>> cognitive-sounding names, and then stick some hand-coded rules or >>> supervised-trained ML models in objects instantiating each of the >>> classes... this may give you a way to architect software programs that >>> are cognitive-ish in information flow and that do stuff, but it won't >>> give you a thinking machine, not by a long shot... >>> >>> ben >>> >>> On Sat, Jul 15, 2017 at 11:17 PM, Ed Pell <[email protected]> wrote: >>> > Holy Cow! They have released this architecture diagram. >>> > >>> > >>> > https://github.com/watson-intu/self-docs/blob/develop/Logical%20View%20(Key%20Classes).jpg >>> > >>> > I wonder how much of that has code. >>> > >>> > -- >>> > 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 post to this group, send email to [email protected]. >>> > Visit this group at https://groups.google.com/group/opencog. >>> > To view this discussion on the web visit >>> > >>> > https://groups.google.com/d/msgid/opencog/d2ad8367-fbc0-438d-8c96-efd4addc3f38%40googlegroups.com. >>> > >>> > For more options, visit https://groups.google.com/d/optout. >>> >>> >>> >>> -- >>> Ben Goertzel, PhD >>> http://goertzel.org >>> >>> "I am God! I am nothing, I'm play, I am freedom, I am life. I am the >>> boundary, I am the peak." -- Alexander Scriabin >>> >>> -- >>> 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 post to this group, send email to [email protected]. >>> Visit this group at https://groups.google.com/group/opencog. >>> To view this discussion on the web visit >>> https://groups.google.com/d/msgid/opencog/CACYTDBdmRHYYRohFWJ%2BTqh-Q0KJYdbctsy36czMYuM%3DznsAQEQ%40mail.gmail.com. >>> For more options, visit https://groups.google.com/d/optout. >> >> > > -- > 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 post to this group, send email to [email protected]. > Visit this group at https://groups.google.com/group/opencog. > To view this discussion on the web visit > https://groups.google.com/d/msgid/opencog/CAHrUA35J0KzdgQPkvPytAay_UdMxitQm7oSCyUqQMEu6OVrJqg%40mail.gmail.com. > > For more options, visit https://groups.google.com/d/optout. -- Ben Goertzel, PhD http://goertzel.org "I am God! I am nothing, I'm play, I am freedom, I am life. I am the boundary, I am the peak." -- Alexander Scriabin -- 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 post to this group, send email to [email protected]. Visit this group at https://groups.google.com/group/opencog. To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/CACYTDBcz4ZGy2KdC5n7QySE18xyJmZfvhfkPOKENSt2ToUz_Cw%40mail.gmail.com. For more options, visit https://groups.google.com/d/optout.
