zlacker

Been a while since I heard about my once-favorite project aiming to imitate common sense. I think I even contributed to its knowledge base a bit. Broken memory is unclear there. I loved that Lenat was one of the few to see (a) the need for common sense representation, (b) many people would need to train it, and (c) good algorithms to integrate it with other things. The part I strongly criticized was locking it up in proprietary fashion: worst thing you can do for stuff needing this much training data.

Good to see it's being commercialized... again? Swore he had a company. Anyway, probably the most valuable thing is the knowledge base they built. It was structured, curated, and very general. It would be great if AI researchers working on different architectures, including adaptive NN's, re-encoded and used that knowledge base. Might speed up training and catch blind spots w/ common sense checks.

Note to other researchers: it would be worth the effort to re-create a similar knowledge base more open to public but with careful moderations. Make sure the knowledge base and decent engine are open source. Gotta be for best results here.

>>nickps+(OP)
> Good to see it's being commercialized... again? Swore he had a company.

My reaction exactly :-)

I got to attend their training in 2003, at Cycorp, which is still around [1]. Some REALLY amazingly smart people.

I wonder if he's saying "it's done!" in hopes of not getting buried by DeepMind... kind of a last-ditch effort for "Strong AI".

[1] http://www.cyc.com

>>chris_+d3
Funny you bring that up. I've been telling people that the one victory we had for common sense was the fact that things like DeepMind approximate it in how they work. They're usually trained on too narrow data vs Cyc. Yet, they do find patterns esp in classification and action/response like human intuition does. I was going to suggest to people that they be used in logical/intuitive hybrids to replace things like Cyc. Maybe hitch a ride on a different bandwagon with benefits from all the effort going into it.

>>nickps+(OP)
It seems terrible that such a project would lock up all that knowledge in a proprietary form.

Fortunately, my scan of their website seems to indicate they have released their ontologies, their under a creative commons license.

http://www.cyc.com/platform/opencyc/ http://www.cyc.com/documentation/opencyc-license/

>>joe_th+G9
Hell yeah! Thanks for the link! Very encouraging for my proposal of integrating it with other methods. :)

>>joe_th+G9
OpenCyc is only a fraction of the ontology, unfortunately. There's a lot of internal desire to update and expand OpenCyc, but my understanding is that at present the company hasn't secured funding that they're really allowed to use for that purpose.

>>catpol+o81
Oh no! I take it back! We're still missing the knowledge base we need. Least OpenCyc might be a nice start on it.