DARPA targets ultimate artificial intelligence wizard

The military's expert research and development arm isn't always about making bigger, better things that blow up or fly fast, sometimes it wants to develop monster brain power. In this case it wants to build avant-garde artificial intelligence (AI) software known as a Machine Reading Program (MRP) that can capture knowledge from naturally occurring text and transform it into the formal representations used by AI reasoning systems.

The idea is that such an intelligent learning system would unleash a wide variety of new AI applications - military and civilian -- ranging from intelligent bots to personal tutors according to the Defense Advanced Research Projects Agency (DARPA).

For example, all of the text in the World Wide Web will become available for automating the monitoring and analysis of technological and political activities of nations; plans, rhetoric, and activities of transnational organizations; and scientific discovery within various disciplines, DARPA stated. As digitized text from library books world wide becomes available, new avenues of cultural awareness and historical research will be enabled. With truly general techniques for effectively handling the incompatibilities between natural language and the language of formal inference, a system could, in principal, be constructed that maps between natural and formal languages in any subject domain, DARPA said.

      DARPA said that  nearly all successful AI systems today succeed because they possess sufficient consistent, relevant knowledge about a given problem. However, since large amounts of knowledge are almost always needed for this success, AI systems require this knowledge to be expressed in a logical formula of some type. Manually encoding such knowledge can become prohibitively expensive. Since text is, by far, the most flexible and ubiquitous medium used to capture knowledge about the diverse areas of human interest, it is natural to consider making it feasible for AI reasoning systems to employ this vast store of human knowledge. As AI systems currently cannot use such knowledge, it would be revolutionary if technology could be developed to bridge this gap, DARPA said.

    The problem is reading and understanding mostly. The necessary information is available, but rarely in a form that can be used by current AI systems, DARPA said. For example, the military frequently faces impediments to stability and reconstruction operations in a new location due to the lack of understanding of the local situation. Similarly, strategic assessment of a foreign nation's science and technology base involves the continuous assessment of technical articles, bibliographies and conference agendas. This information is often available on the Web, and some tools to assist this analysis are available, but the process would be significantly enhanced by a system that could directly analyze the information found in these text sources. The same reasoning could be equally valuable if applied to other types of open-source intelligence analysis, including assessing military readiness and posturing; political speeches, actions, and more obscure messages; economic trends and sentiments; and propaganda from terrorist groups and even their hidden web-based communications.

    DARPA lists some the technological goals of the new AI system as follows:

  • A Universal Reading System: Create a universal Reading System that can take any natural text and any reasoning context as input and can effectively apply the knowledge contained in that text in that reasoning context.
  • Enhanced Capabilities from Combining natural language and AI reasoning: Combine

NLP and AI reasoning into new technology that provides the benefits of both.

  • Domain-Specific Performance from a General-Purpose System: Develop a general-purpose text-reading "front-end" (Reading System, for short) that can be used with any number of domain-specific reasoning systems (DSRS).
  • Couple NLP and Convergent Knowledge: The goal in coupling NLP with convergent knowledge - knowledge that is mutually constraining - is not to understand facts implied by a single sentence in isolation. Rather, the goal is to uncover physically remote facts, contexts and determine how these bits of knowledge can, in turn, help make the NLP feasible by constraining the interpretation of other such bits of information.
  • Deployment of NLP in a Wide Range of Practical Reasoning Contexts: Develop easily reused and re-purposed NLP technology that can be rapidly applied in a wide range of practical reasoning contexts where textual knowledge could provide strong value.
  • Research Community in Machine Reading: Another important goal is to nurture a research community that focuses on the machine reading problem.

    How such software will ultimately be contracted and developed will be big issues themselves. Some of the requirements are extensive.

    DARPA has been interested in exploiting the promise of AI for years.  Earlier this year it approved the second phase of artificial intelligence technology that will help automate military air traffic control.   The Generalized Integrated Learning Architecture (GILA) system, developed by Lockheed Martin's Advanced Technology Laboratories under a $22 million, 48-month contract, is intended to help the Air Force in particular keep airspace operating safely with increased air traffic and the advent of unmanned aerial vehicles (UAVs) and other airborne weapons.

Layer 8 in a box

Check out these other hot stories:

Telemarketing poppycock elimination laws take effect today

NASA goes iTunes with free podcasts

FAA  greenlights satellite-based air traffic control system

Top 10 wicked cool algorithms

IBM supercomputers greenest, fastest

Wal-Mart blows into wind power in a big way

Jack of all dark trades: Swatter, botnet herder, hacker pleads guilty

NASA exploring 8 new space expeditions

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.
Related:

Copyright © 2008 IDG Communications, Inc.

IT Salary Survey: The results are in