Matt, I agree that Cyc is a non-workable approach to AGI, however I would note there is in fact a sophisticated NL interface to Cyc (and has been for a very long time).
I also note that there have been Bayesian learning algorithms associated w/ Cyc for decades now ... which work together w/ their predicate logic theorem-proving engine. Again I don't like the overall approach with its focus on hand-coded knowledge, and it also seems they don't have any scalable approach to inference control... but there is in fact a learning algorithm and NL interface, unlike what you say... Stephen Reed's CognitiveCyc architecture explained how to put experiential reinforcement learning together with Cyc, also. But then he quit Cycorp to develop this approach in a separate company, Texai was the name last I checked. I have not kept current on his work. ben On Sun, Apr 11, 2021 at 6:05 PM Matt Mahoney <[email protected]> wrote: > > Cyc is a lesson on why symbolic AI doesn't work. Its major shortcomings are > that there is no learning algorithm and no natural language interface. How do > you address these issues? > > On Sun, Apr 11, 2021, 4:31 PM stefan.reich.maker.of.eye via AGI > <[email protected]> wrote: >> >> For your reading pleasure. >> >> https://botcompany.de/files/1400465/Gazelle+AGI+Technical+Pitch.pdf > > Artificial General Intelligence List / AGI / see discussions + participants + > delivery options Permalink -- Ben Goertzel, PhD http://goertzel.org “He not busy being born is busy dying" -- Bob Dylan ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T2e6144b49eda1e2e-Md672e0a62e31a8b814ae323d Delivery options: https://agi.topicbox.com/groups/agi/subscription
