From: Ben Goertzel •t ), To: Jeffrey Epstein <jeevacation@gmail.com> Subject: Re: Brainfood... Date: Sat, 02 Apr 2016 14:31:10 +0000 And a fuller project plan... https://docs.google.com/document/d/IwzpMCQmDIGT4tDCKetnXbNx-ticMCICwIs_Z-E4UNmego/edit On Thu, Mar 31, 2016 at 5:39 PM, Ben Goertzel wrote: > And a bit more project-planning stuff for our envistone -year > OpenCog initiative (courtesy of Cassio) > On Tue, Mar 29, 2016 at 11:37 PM, Ben Goertzel •t > wrote: » Hi Jeffrey, >> So, here is some food for thought and discussion... >> We are working on getting OpenCog to control the Hanson Robotics robot >> heads for embodied (or at least "emheadied"!) natural language >> dialogue. This should be working in Summer 2016, maybe sooner. We >> hope to demonstrate this to some extent at AGI-16 which will be in New >> York this July. To give you some flavor of this work I attach our >> Hanson Robotics / OpenCog software plan for the next few months (this >> is work funded by Hanson Robotics). >> In parallel with that, we have been finally getting some examples of >> cognitive synergy to work in OpenCog. >> This paper of ours from AGI-14 ended with a bit of a whimper >> http://agi-conf. org/2014/wp-content/uploads/2014/08/harrigan-guiding-agi14.pdf >> however we have recently actually gotten this to work. I.e. we have >> gotten ECAN (neural net type activation spreading) to successfully >> prune the PLN inference tree. If all goes well we will submit a paper >> on this to AGI-16 next week. This shows attention allocation and >> inference working together. Yay! » This paper of ours from AGI-15 shows PLN being used to estimate the >> truth value of a pattern recognized in data by MOSES (our evolutionary >> program learning algorithm) >> >> http://agi-conforg/2015/wp-content/uploads/2015/07/agi15_goertzel_speculative.pdf >> It takes about 24 hours to run in the current OpenCog version. >> Obviously this is not OK but it's a start. >> Next step is to put these two things together, and use ECAN to help >> PLN inference pruning on nontrivial examples like the one from the » above AGI-15 paper. EFTA00830265 >> The attached PPT on some of our biology work contains two parts >> -- first part is "straightforward" Al-wise but has quite interesting >> results biology-wise: i.e. we used MOSES to figure out genes that >> distinguish supercentenarians from healthy 80 year olds >> >> -- second part is more interesting AI-wise ... using PLN to generalize » from MOSES models >> » Between these papers we see the first concrete examples of "cognitive >> synergy" in OpenCog. They are a bit "toy" but they are also real and >> actually work in the codebase now ;) >> >> Finally, the other PPT attached shows the plan that Jim Rutt and >> Cassio and I are aiming to carry out over the next 3 years. I am >> not sure if you intersected with Jim at SFI or not... >> >> http://www.santafe.edu/about/people/profile/Jim%20Rutt >> »We are aiming to raise $6M for 3 years of work, but I imagine we could » do it for $4M over 3 years if we made copious use of inexpensive >> overseas folks. Jim will put in at least $500K of the funds himself >> (his foundation gives out $500K/year, and he is willing to put at » least 1/3 of that into OpenCog, maybe more). >> >> The goal of this plan is to make OpenCog-based embodied NLP dialogue >> work extremely well and hold a conversation demonstrating obvious >> understanding of itself and its environment and situation. At the >> core this will require getting cognitive synergy between ECAN, MOSES » and PLN to work in the context of visual/auditory perception and NLP >> dialogue. This will be both a huge AGI advance and an amazing demo >> that will open pretty much any door we want. >> >> Jim Rutt was a good find, not only because he has been donating a bit >> of $$ (about $150K last year, none to me personally but he has funded >> Nil Geisweiller and Cassio), but because he bring a management and >> organization and business sensibility/experience that complements me >> well. And he knows enough science and CS and has put in enough time >> to thoroughly understand the OpenCog/CogPrime design... >> » thx! » Ben » -- » Ben Goertzel, PhD » http://goertzel.org > -- > Ben Goertzel, PhD > http://goertzel.org EFTA00830266 > > "I am Ubik. Before the universe was, I am. I made the suns. I made the > worlds. I created the lives and the places they inhabit; I move them > here, I put them there. They go as I say, then do as I tell them. I am > the word and my name is never spoken, the name which no one knows. I > am called Ubik, but that is not my name. I am. I shall always be." -- > Ubik Ben Goertzel, PhD http://goertzel.org "I am Ubik. Before the universe was, I am. I made the suns. I made the worlds. I created the lives and the places they inhabit; I move them here, I put them there. They go as I say, then do as I tell them. I am the word and my name is never spoken, the name which no one knows. I am called Ubik, but that is not my name. I am. I shall always be." — Ubik EFTA00830267