«  
  »

Cyc project (partially) open-sourced

By RichardTerra, on June 11th, 2002

WillWare writes "There has recently been released a partially open-sourced version of the Cyc artificial intelligence program. Information can be found at http://www.opencyc.org/. Cycorp, the commercial venture that developed Cyc, writes the following on the benefits of this release.

From this point forward, real-world common sense can be expected to play an integral part of software applications. For the first time, the world's only large-scale, task-independent, language-independent, extensible, reusable, common-sense knowledge base is being made available to the world. Beginning now, software can become increasingly and arbitrarily smarter.

They expect the following benefits for themselves.

To establish Cyc as the standard for knowledge representation, for knowledge management, for data base integration, and in general for intelligent software applications. Also, the release of OpenCyc will help lay the groundwork for the massively parallel effort to rapidly grow the Cyc [knowledge base]… OpenCyc will raise awareness for symbolic knowledge representation. It will also create opportunities for combining symbolic and rule-based systems with other technologies, such as neural networks, planning systems, machine learning and genetic algorithms.

One might expect that the most MNT-relevant AI programs will look more like CAD programs than common-sense databases, at least in the near future. Where the Cyc technology would have an advantage would be in addressing problems that are not already well formulated and well studied. Perhaps a Cyc-based program could be a good researcher, or a good economic or political advisor, or would fare better in meeting the unexpected challenges that might be encountered on a long term space mission than a more clearly goal-specific program. In any event, it's an interesting development, and it will be intriguing to see what is done with the open-source release."

June 11th, 2002 | Category: Machine Intelligence, News

3 comments to Cyc project (partially) open-sourced

  • Mr_Farlops
    June 15th, 2002 at 6:15 PM

    Re:Helpful for commercial AI

    I didn't mean to say that Cyc was old-fashioned. It is still very relevent for commercial applications–database searching technology, demographic analysis, perhaps even network design–things that need to make non-intuitive associations.

    I was just saying that it's irrelevent to research in strong AI. I think it is more likely that strong AI will emerge from some combination of computational neurology and robotics–which attempt to reverse engineer the essential aspects of the brain, senses and body to produce consciousness.

    But in commercial applications, all we are worried about is formalizing mental processes used by experts. Consciousness doesn't enter into it.

    An interesting questions raised by Cyc are:

    Can we formalize intuitive leaps in software? Can we learn from Cyc to create a formal version of the mental processes that Ramanujan used to come up with all his bizarre, yet mostly correct, mathematical formulae? This could be very useful.

  • WillWare
    June 14th, 2002 at 9:19 PM

    Re:Helpful for commercial AI

    If Cyc is no longer relevant (quite believable, as it's by now fairly long in the tooth) what's closer to something current? And in particular, has anything else interesting been released as open source?

    I see Sourceforge has something called OpenAI, and I'm aware of FramerD, but neither of them include a large ready database of common sense concepts, rules, and associations. I also found an open-source section of the Usenet AI FAQ that seems to have some interesting stuff. Again, plenty of algorithms but no ready collections of data. The Alicebot project has some data, but if the example is any indication, the data may not be terribly interesting or general-purpose.

    Eli Yudkowsky has plenty of interesting ideas but I am not aware of any software yet.

  • Mr_Farlops
    June 13th, 2002 at 10:42 AM

    Helpful for commercial AI

    This will probably be quite helpful for various commercial applications of AI.

    However I think cyc really isn't that relevent to research in strong AI anymore. It's sort of like Ted Nelson's Xanadu Project, years in the making with many great claims made–meanwhile the Web, which is admittedly a modest implementation of Xanadu, has long since moved on.