|
COPYRIGHT 2006 American Association for Artificial Intelligence
This article is derived from the 2005 Innovative Applications of Artificial Intelligence conference invited talk "AI Meets Web 2.0: Building the Web of Tomorrow, Today" presented by Jay M. Tenenbaum in Pittsburgh, Pennsylvania. I invited Marry to deliver that talk at IAAI-05 because he is a true AI and web visionary. It is common in times of grade inflation and hyperbole that speakers and authors are introduced as "visionaries." Jay M. Tenenbaum (or Marty, as he is known) actually is a world-renowned Internet commerce pioneer and visionary. Mary envisioned the commercial and industrial use of the Internet more than a decade before it became a reality, at a time when only academics, industrial R&D groups, and government labs had access to it. At the time, Marty's e-commerce vision evoked all kinds of resistance and objections about why it couldn't or shouldn't be done.
In 1991, Marry founded and became the chief executive officer of Enterprise Integration Technologies, the first company to conduct a commercial Internet transaction (1992), a secure web transaction (1993), and an Internet auction (1993). In 1994, Marty formed the CommerceNet industry association and jump-started the Internet marketplace by convincing 50 leading corporations to get online. In 1997, he cofounded Veo Systems, which pioneered the use of XML documents for automating business-to-business transactions. In 1999, Commerce One acquired Veo Systems, and Marty ultimately served as its chief scientist. Marry now splits his time between Internet startups Efficient Finance, Medstory, and Patients Like Me, and the nonprofit CommerceNet.
This article points to the technical opportunity to meet Allen Newell's criteria for intelligent systems by incrementally assembling complex web services applications out of component web services, tagged data, and inference techniques. It also articulates an exciting set of business opportunities to deliver on the economic promise of e-commerce and AL
- -Neil Jacobstein, Chair, IAAI-05
* Imagine an Internet-scale knowledge system where people and intelligent agents can collaborate on solving complex problems in business, engineering, science, medicine, and other endeavors. Its resources include semantically tagged websites, wikis, and blogs, as well as social networks, vertical search engines, and a vast array of web services from business processes to AI planners and domain models. Research prototypes of decentralized knowledge systems have been demonstrated for years, but now, thanks to the web and Moore's law, they appear ready for prime time. This article introduces the architectural concepts for incrementally growing an Internet-scale knowledge system and illustrates them with scenarios drawn from e-commerce, e-science, and e-life.
In this article, I want to share a vision of how to build or, more precisely, grow Internetscale knowledge systems. Such Systems enable large numbers of human and computer agents to collaborate on solving complex problems in engineering, science, and business or simply managing the complexities of life (say planning a trip or an event). It's a vision that's been evolving over 20 years since my days as an AI researcher and, more recently, as an Internet entrepreneur. Thanks to the explosive growth of the web, it's a vision whose time has come. I also have a larger goal: to bridge the AI and web communities, which have so much to give to and learn from each other.
Twenty-five years ago, at the birth of AAAI, Allan Newell articulated a set of criteria that a system had to exhibit to be considered intelligent (see table 1). Newell was very explicit that an intelligent system had to exhibit all of these criteria. This requirement reflected the then prevailing view that intelligent systems were monolithic and were developed centrally by an individual or small group.
Table 1. Newell's Criteria for Intelligent Systems. * Exhibit adaptive goal-oriented behavior * Learn from experience * Use vast amounts of knowledge * Exhibit self-awareness * Interact with humans using language an speech * Tolerate error and ambiguity in communication * Respond in real time
The web has shown us a different path to intelligence--millions of simple knowledge services, developed collaboratively in a decentralized way by many individuals and groups, all building on each other, and demonstrating a collective form of intelligence. It's time to reconcile these two views and create a new breed of hybrid knowledge systems that combine the best elements of AI and the web, and of humans and machines. Such hybrid systems can solve heretofore intractable real-world problems and potentially demonstrate heretofore unattainable levels of machine intelligence.
To give a hint of where I'm headed, I'll revisit some classic AI problem-solving tasks, which are as timely now as they were in Newell's day (table 2). I've worked on many of them, and I'm sure readers have too. How would you as an individual approach any of these tasks today? Undoubtedly, by searching the web, which provides thousands of relevant information sources and services.
Table 2. Classic AI Problems. Travel What's the best way to get to Pittsburgh? Meetings Who should I have dinner with tonight? Where? Supply chain How can I get 100 PCs delivered by tomorrow? Medicine What drugs might be effective against this new bug? E-business Where can get the best deal on a car and a loan?
If I were building an AI system to solve these problems, I too would start with the web, gradually adding a little AI here and there to automate time-consuming functions and integrate processes. Take travel, for example. I might start by creating a simple agent to check many travel sites for available flights and fares. A second agent might use these results to select an optimal itinerary based on my personal preferences, frequent flyer accounts, and calendar schedule.
Now imagine giving millions of people this ability to create agents that incrementally automate tasks, as well as the ability to publish those agents as services that others can use and build upon. Before long, interesting forms of collective intelligence will begin to emerge.
Approaching AI applications in this way has manifest benefits--the ability to leverage billions of dollars of web infrastructure, millions of web services, and millions of knowledge engineers all building on each other's work. Moreover, such systems are useful from day one, because they start with the web. Automation proceeds incrementally--one person, and one task at a time. Each step provides immediate, incremental benefit. And because people remain in the loop, such systems can fail gracefully, by handing off hard or unusual cases for manual processing.
I'll start this article by reviewing the evolution of the vision from its roots in AI research through the early days of the web and more recent developments such as web services and Web 2.0 (referring to the spate of recent web innovations that includes blogs, wikis, social networks, and the like). These recent developments make the web more programmable and more participatory, important attributes for Internet-scale knowledge systems. I'll review Web 2.0 technologies and methodologies from an AI perspective and then illustrate, through case Studies and scenarios, some knowledge systems that synthesize Web 2.0 and classic AI approaches. Finally, I'll conclude with a call for action for realizing the vision on an Internetwide scale.
Evolution
Twenty years ago, I was running a research lab for Schlumberger. At the time, Schlumberger was trying to break into the CAD/CAM business through its acquisition of Applicon. I was fascinated by the possibility of using the Internet to support large-scale engineering projects, such as the design of a new airplane, involving thousands of people at hundreds of companies (figure 1).
[FIGURE 1 OMITTED]
Inspired by distributed blackboard systems such as Hearsay, we modeled the process as human and computer agents collaborating through a shared knowledge base, representing a model of the artifact. When an agent modified the design, affected agents were notified so they could critique the change or respond with changes of their own. For example, a hydraulics engineer installing some fluid lines might alert an airframe agent to check whether those lines interfered with the movement of any control surface. Although a centralized shared knowledge base is depicted, the model can be distributed in practice to facilitate scaling. Each agent maintains aspects of the model most relevant to it in local CAD systems and provides that information to other agents that need it.
The Palo Alto Collaborative Testbed (PACT)
We built several research prototypes of such agent-based collaborative engineering systems during the late 1980s. The most impressive was the Palo Alto Collaborative Testbed (PACT), which was documented in the January 1992 issue of IEEE Computer. PACT used the Internet to link four independently developed knowledge-based design systems (covering requirements management, kinematics, dynamics, and electronics) at three different...
Read the full article for free courtesy of your local library.
|
Bookmark this article
Print this article
Link to this article
Email this article
Digg It!
Add to del.icio.us
RSS
