GoogleAI--Adding Commonsense to Google

[Christopher McKinstry]

I am very happy to see the webapi finally announced. I think it's a
very smart move.

For those of you who have no idea who I am, I run the world's largest
artificial intelligence project http://www.mindpixel.com

I've been thinking for some time about dumping my database of almost
500,000 human created/human validated commonsense facts into webpages
so that Google would index them... that way, systems could make
commonsense propositional queries to Google... kinda like copying a
whole person into the Google index.

Each page would have the text of the proposition, plus some phonetic
and semantic transformations of each of the base propositions, plus a
human validated probability of being true. The transformations would
allow for soft hits against the text of the propositions... for
exapmple if you asked google "Am I a conscious human being" a simple
soundex transform to "m cns hmn bn" would allow for many various
misspelling to hit the fact and return a true.

I know I'm babbling... I hope readers of this understand what I'm
proposing... to make Google not only a search engine, but a
commonsense search engine.

I love to do this tomorrow, but I have finicial obligations to the
people who built the Mindpixel corpus... does anyone know who I would
contact at google about licensing our corpus to Google? Such a deal
would be good for the mindpixel project, google and all the webapi
developers.

And, yes I know the current Mindpixel interface sucks... it's still
running 2 year old code because I'm alll alone down here in South
America without any development support... something else that a
Google deal would change.

Thanks for reading this far...

Chris

ch...@mindpixel.com
cmck...@eso.org
(if you send me mail, copy both addresses)

[Oblom]

Hey Chris,

> For those of you who have no idea who I am, I run the world's largest
> artificial intelligence project http://www.mindpixel.com

Just wondering if you heard of opencyc.org. Maybe you could
incorporate your KB with theirs.

Best,

Oblom.

[Christopher McKinstry]

> Just wondering if you heard of opencyc.org. Maybe you could
> incorporate your KB with theirs.

Sure, I know about opencyc--Doug Lenat and I both have chapter's on
our respective projects in a forthcoming book on the Turing Test from
Kluwer Academic Publishers in Holland due out next year...

And don't forget, there is also MIT's openmind project. They also have
about 500,000 propositions in the public domain... mind you, their
dataset is noisy as no one validates anything in it. I plan on pulling
in both opencyc and openmind into Minpixel in the near future, where
their propositions can be enhanced by my user population. That would
give me a corpus of well over 1,300,000 facts, 500,000 of which are
actually measurements of human propositional hyperspace, and not just
"facts".

I'm quite excited about this GoogleWebAPI thing and Google in
general... I think the future will see Google become the internet! I
already prefer to use Google's cached results instead of going to the
source site, becasue my search terms are highlighted, and it's
generally much faster for me here in South America...

Now, if we enhance Google with commonsence, then Google becomes not
just a copy of the internet, but a copy of humanity--with an API!

Chris.

BTW: I'm looking for developers to help me with the whole system...
I've been concentrating on the theory, and neglecting the system. I
really need help!!! Email me directly at cmck...@eso.org if
interested.