A Breath of Fresh Air: Project Oxygen

Dateline: 09/28/00

After more than 50 years of development and technological breakthroughs, we'll have to face the truth: computers are still too difficult to use. If you are reading this article, chances are you have considerable experience and know a thing or two about computer hardware, various operating systems and Web browsers, so you may not agree with me. But from a perspective of an absolute beginner, the whole process of interaction with modern computer equipment looks quite confusing. You don't even have to be a beginner: I just spent several days trying to quickly migrate to the newest release of a popular server operating system. It wasn't a job I would like to do every day. Does it have to be so difficult and why can't we communicate more naturally with our machines? Where did "pervasive" and "human-centered computing" go?

Several AI-related technologies come to the rescue. Contrary to the industry cliche that Artificial Intelligence did not fulfilled people's expectations, I believe it has succeeded as a part of a big picture - technology buried inside other tools. Same philosophy is followed by the new project initiated at MIT and recently funded with $50 million from six major companies and U.S. Defense Advanced Research Projects Agency (DARPA). Joining the Oxygen Alliance this June were Acer Inc., Delta Electronics Inc., Hewlett-Packard Co., Nippon Telegraph and Telephone Corp., Nokia Corp. and Philips Research, an arm of Philips Electronics. The primary goal of the Project Oxygen is to develop computers that will be freely available everywhere, like oxygen in the air we breathe. This new generation of computing devices will enable people "to do more by doing less", that is, to accomplish more goals with less manual and routine work.

The project is a brainchild of Michael Dertouzos, director of MIT Laboratory for Computer Science, and is pursued in collaboration with the Artificial Intelligence Laboratory led by Rodney Brooks. It employs eight technologies - 3 hardware and 5 software - that enable natural use and increased productivity. Starting with hardware gadgets, future Oxygen users will carry only a handheld portable computer, currently about the size of two paperback books. This portable universal device, code-named Handy 21, will combine functionality of today's cell phone, two-way radio, TV, beeper, handheld computer, pointing device, GPS and more. The main point here is software controllability: except for a tiny analog part connected to the antenna, the behavior of every part can be finely tuned by the user and software components working for him. Next is Enviro 21, out-of-sight unit embedded into walls, ceilings, car trunks and similar places. Basically, it has the same functionality as Handys, but with greater processing power and communication speed. Enviros will also be used to control sensors, actuators and standard home appliances. The N 21 network links all Oxygen devices to each other and to the Internet. It will self-organize and adapt to the various environments beyond the capabilities of current Internet protocols, which is necessary to keep up with users moving from place to place and using various general-purpose devices. Finally, at the hart of the MIT's concept of ubiquitous computing is a new, specialized chip developed as a part of Raw (Reconfigurable Architecture Workstation) project.

Other user technologies provide additional functionality needed for stretching the limits of human-computer interaction. The individual knowledge-access technology helps you find the information you need from heterogeneous data stores, using a number of cooperating softbots. It adapts to your needs and automatically escalates simply-phrased queries, from local hard drive, via your friend's databases to the Internet. The automation technology lets you offload the routine work to the computer, by simply writing scripts for various boring jobs. The collaboration technology helps controlling a groupwork by presenting summaries of previous speech fragments and documents, tracing the discussion and keeping a trail of all previous issues. The customization technology adapts Oxygen to the needs of individual users, automatically downloading new versions of the software when they become available. Conversational input-output is at the core of the project: in involves not only speech recognition and generation, but also language understanding technology. The Jupiter Web site will give you a glimpse of what's to come. Of course, current applications are domain-specific and limit conversation to a narrow manageable topical area. But who knows what can we expect in the next 5 years or so...

A working Oxygen prototype is scheduled to be built later this year. More good news: all of the parties working on a project have agreed that the software will be open source. Should we throw away those keyboards? Not really, as the first "real-world" applications are expected for 3 to 4 years. However, as Mr. Dertouzos says, this project should encourage application developers and users to bend machines toward human needs, rather than the other way around. Don't think that this project is one-of-the-kind: other industry giants like IBM and Intel have similar efforts underway. DARPA is also funding four other university contractors on the "ubiquitous computing" programs. Keeping both feet on the ground will also help defying those old AI-related cliches I was talking about at the beginning of the article. "We don't make intelligent or smart anything", says Dertouzos, but the final result could be smarter that the sum of its parts.