After checking the http://old.reddit.com/r/OpenCog subReddit for the past year or so, yesterday I had the pleasant surprise of seeing that "nickb" had posted an announcement twenty hours earlier about an upcoming on-line conference. Searching for more information, I came across http://multiverseaccordingtoben.blogspot.com/2020/05/gtgi-general-theory-of-general.html as a thought-provoking blogpost by BenG (Ben Goertzel). Late at night I captured a local copy of the text so that I could write this response in the morning.
In the second paragraph, BenG makes an intriguing mention of "consciousness from first, second and third person perspectives", which immediately reminded me of the http://ai.neocities.org/var.html#pov variable which I use in my various AI Minds to indicate a "point of view" (POV) which can refer to first, second or third person. I am pleased to see BenG in the fifth paragraph referring to a "new conceptual model of mind", because I feel that AGI software needs to include concepts. In paragraph six BenG talks about "psi phenomena" -- which I shy away from -- although I use "Psi" as the name of the conceptual array in my AI Minds. Paragraph eleven mentions Occam's Razor, which always perks my interest because I long ago memorized its statement in the original Latin: "Entia non sunt multiplicanda praeter necessitatem" -- "Entities are not to be muliplied outside of necessity." Paragraph thirteen alludes to Charles Saunders Peirce, a name known to me because an erstwhile acquaintance of mine, about whom a whole chapter was written in "Mathematical Cranks", was often talking about Peirce when we sat together in the long-gone "Last Exit" coffee shop on Brooklyn Avenue near the University of Washington. Oh, gee. Paragraph fourteen mentions Donald Knuth, whom the mathematical crank was often talking about. In paragraph 18 BenG says that "one can look at intelligence as the ability to achieve complex goals in complex environments using limited resources." Here is where I must diverge from BenG's approach to AGI. Instead of worrying about goals for a living organism, since 1993 I have always tried to build artificial Minds by creating first concepts and then associative interactions among concepts. Paragraph 21 talks about "CoDDs, Combinatorial Decision Directed-acyclic-graphs" as a "computational model", and here I must assert that the Mentifex approach to building AI Minds that think in English, German, Latin or Russian is based on the gradual accretion (adding on) of software implementations of concepts and their associative interactions under the guidance of Natural Language Processing (NLP), a coin whose obverse side is Natural Language Understanding -- http://ai.neocities.org/NLU.html -- (NLU). Paragraph 32 talks about the Cognitive Synergy "which underpins the OpenCog AGI design". Here I would like to suggest that Netizens who hunger and thirst for TrueAGI should build elementary Minds out of concepts and NLP structures, then let such dawn-of-AGI entities compete a la "survival of the fittest" to evolve all the more advanced goal-seeking mechanisms. I mean, we can create software that thinks, then let the thinking software evolve into advanced goal-seeking software. Oh! Paragraph 34 accords with the Mentifex approach inasmuch as BenG says that "These specialized cognitive algorithms must be learned/evolved based on multiple constraints...." In paragraph 35 BenG describes the "chicken-egg problem -- which is solved by human AGI system designers performing the first round of the iteration." I want the Menfifex AI Minds to be the egg which evolves into the AGI chicken. https://www.mail-archive.com/agi@agi.topicbox.com/msg05555.html https://www.mail-archive.com/agi@agi.topicbox.com -- list of messages. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T295a3bd072ac1b6f-Mb8a9e23179c5107da6e53b48 Delivery options: https://agi.topicbox.com/groups/agi/subscription
