See also Arthur Franz's work on biasing AIXI (Hutter's infinite-resources theoretical totally general AGI) toward our physical universe. He gave a great talk on this at AGI-19 in Shenzhen in August, and the video may unfortunately not be online yet, but I think he has a PDF paper on this posted somewhere also... (he has an AGI team working out of Kiev...)
ben On Tue, Feb 4, 2020 at 7:24 PM Danko Nikolic <[email protected]> wrote: > > Hi Ben, > > Thanks for that information. Let me mull over it. > > Danko > > On Tue, Feb 4, 2020 at 9:57 AM Ben Goertzel <[email protected]> wrote: >> >> Are you familiar with Marcus Hutter's work on Universal AI? >> >> The general message there is: Truly, fully general intelligence is >> only feasible with infinite computing power. >> >> However, that's not really what matters in our physical universe, let >> alone in the human world here on Earth... >> >> Eons ago I aimed to formalize a more real-life-oriented variant of >> Legg and Hutters formalization of the general intelligence notion, >> >> http://agi-conf.org/2010/wp-content/uploads/2009/06/paper_14.pdf >> >> The "no free lunch theorem" mathematically connects w/ the statistical >> decision theory underlying Hutter's work, it's all essentially the >> same story. In essence, we may not get a free lunch but we can get a >> cheaper lunch if we buy it from our relatives. I.e. yes real-world >> general intelligence has to do with biasing somewhat-general >> intelligence toward the specific patterns in the actual universe the >> intelligence has to deal with. Deep NNs are a small example of this, >> in that their hierarchical structure represents a biasing toward the >> hierarchical structuring commonplace in our region of the physical >> universe. >> >> I tried to formalize the relation btw world/goal structures and >> mind-structures here, >> >> https://www.semanticscholar.org/paper/A-Mind-World-Correspondence-Principle-Goertzel/ec4ca4401f38ec10d8d3cf40c39af0a4bfe86557 >> >> How useful this sort of formalization is for actually building AGI is >> an open question. My current AGI research is focused on >> formalizations that bridge the math used in practical integrative AGI >> systems like OpenCog with the math used in these sorts of abstract >> formalizations... >> >> -- Ben >> >> On Mon, Feb 3, 2020 at 8:44 PM Danko Nikolić <[email protected]> >> wrote: >> > >> > Hi, >> > >> > I am curious to learn what the people here think about the idea that the >> > notion of general intelligence contradicts the no-free-lunch-theorem. >> > >> > Do you guys agree that a contradiction exists? If so, is that a problem? >> > Would that mean that no form of general intelligence could ever exist? >> > >> > Also, does anyone here undertake any explicit efforts to circumvent this >> > contradiction? >> > >> > Thanks a lot. >> > >> > Danko >> > >> > Artificial General Intelligence List / AGI / see discussions + >> > participants + delivery options Permalink >> >> >> -- >> Ben Goertzel, PhD >> http://goertzel.org >> >> “The only people for me are the mad ones, the ones who are mad to >> live, mad to talk, mad to be saved, desirous of everything at the same >> time, the ones who never yawn or say a commonplace thing, but burn, >> burn, burn like fabulous yellow roman candles exploding like spiders >> across the stars.” -- Jack Kerouac > > > > -- > Prof. Dr. Danko Nikolić > www.danko-nikolic.com > https://www.linkedin.com/in/danko-nikolic/ > Artificial General Intelligence List / AGI / see discussions + participants + > delivery options Permalink -- Ben Goertzel, PhD http://goertzel.org “The only people for me are the mad ones, the ones who are mad to live, mad to talk, mad to be saved, desirous of everything at the same time, the ones who never yawn or say a commonplace thing, but burn, burn, burn like fabulous yellow roman candles exploding like spiders across the stars.” -- Jack Kerouac ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T353f2000d499d93b-M2f84c45164bddfb6b6a78151 Delivery options: https://agi.topicbox.com/groups/agi/subscription
