Of all the semantic web projects and technologies, perhaps none is bigger than CALO, a so far little-known advanced intelligent assistant system with some big-time backing.
CALO stands for Cognitive Assistant that Learns and Organizes, and it emerged from an ambitious program of artificial intelligence (AI) research. One expert called CALO the "largest AI project in history."
It's very important to DARPA, the Defense Advanced Research Projects Agency, which is investing in it heavily. About 25 universities and companies are working on CALO, including MIT and USC; the companies include PARC and Radar Networks.
CALO was demonstrated in detail Thursday night in San Francisco at a meeting of the SDForum, an eclectic group of Bay Area entrepreneurs, programmers, and those interested in the semantic web.
CALO -- which semantic web pioneer Tom Gruber calls one of the most "pure play examples" of an intelligent assistant -- learns about your documents, email, people, schedules, and meetings, and learns even more as you use it. It helps you organize your information world, prepare for meetings, create presentations, and find information in the context of your work.
Adam Cheyer demonstrated some of CALO’s powers Thursday night. Cheyer is a scientist with an impressive rap sheet, the kind developed after many years in Silicon Valley. He is a program director in SRI's Artificial Intelligence Center, where he serves as chief architect of the CALO/PAL project. Cheyer is also senior scientist and co-director of the Computer Human Interaction Center (CHIC) at SRI International.
Cheyer said that CALO is being developed in an office format, but it's being transitioned to the military for various projects like "command post of the future" and "PlatoonLeader."
CALO has three main high-level functions: information management, meeting understanding, and task management execution.
Cheyer said that other products have done meeting understanding, but CALO is different and more robust because, for example, it knows things like who is in the meeting, what the people do, and what documents are important to this meeting. If CALO notices that a certain manager -- let's say Manager Adam -- isn't in a meeting, it starts to ask questions about Adam. Maybe he isn't really the project lead? It then does "machine learning" based on those questions.
To see how CALO is doing at a particular company, CALO’s creators have set up a very detailed way to test the system.
Here's one example: Let's say Executive Smith had a CALO program running for two weeks, and that program was supposed to be learning all about Smith and all his obligations, professional contacts, presentations, emails, calendar appointments, etc. during those two weeks.
CALO would then know all about the meetings Smith attends, and, if Smith can't make a meeting, CALO would automatically suggest who should attend the meeting in Smith's place, and even email the replacement.
IE 7
Firefox 2.0
