Reading Arpa's research entrails to determine the future direction of technology.
Arpa, the Advanced Research Projects Agency of the US Department of Defense, is best known for funding the development of the Internet. But their monetary support has also brought us important technologies such as RISC microprocessors and flat-panel displays. Although intended for military use, these devices have thoroughly infiltrated our daily lives. To find out what's ahead, it makes sense to look at what Arpa is funding today.
Unfortunately, that isn't easy. Arpa's scattershot approach to funding supports lots of small and diverse projects rather than a few big ones. So, to get some sense of the hot topics and how they interrelate, Wired mapped Arpa's research using a technique known as co-word analysis. Originally developed by sociologists studying the spread of scientific ideas, the procedure exposes the forces and structures embedded in text.
First, we analyzed all the Arpa project summaries related to computer technology and picked out the most common technical keywords, such as network and imaging. Then we mapped the results: words that commonly occur together in project descriptions are located near one another, and the type size of a word reflects its frequency. Linked keyword pairs are connected by lines, whose thickness indicates the strength of the connection. For example, the final map shows a thick line between ATM and network because almost every project that mentions one of these words also mentions the other.
The map exposes two main clusters of research. On the left, the focus is on parallel computing. Words like compiler, language, and memory encircle parallel, reflecting the key concerns of the field. On the right is the network cluster, largely unlinked to any of the terms that surround parallel. Here, applications such as imaging and encryption, as well as technologies like ATM and mobile, radiate out.
Co-word maps are an efficient way to visualize the structure of a research field. Of course, they also have their pitfalls. They don't distinguish between multiple meanings of a word, for instance, which can produce misleading results. But as a first approximation, a co-word analysis provides a useful battle chart of scientific research and a peek at the future.
It's no surprise that mobile and security lie so close together. A portable computer is far more likely to end up in prying hands than a computer tucked away in an office. To prevent tampering, scientists at Carnegie Mellon University have built a secure coprocessor that can be entrusted with the user's digital key and can compute software checksums to guard against viruses and Trojan horses. www.cs.cmu.edu/afs/cs.cmu.edu/project/coda/Web/coda.html
Arpa researchers haven't forgotten their early creations. A few people at Bolt Beranek and Newman Inc. are studying how to add encryption to the Internet. The proposed scheme allows documents to be digitally signed, encrypted, and time-stamped, then accessed with standard Net software such as mail programs and Web browsers.
The biggest problem with parallel computers it that they are hard to program. The solution: smarter compilers and simpler languages such as High Performance Fortran. www.npac.syr.edu/hpfa/
Do we really need parallel computing? Arpa's efforts make clear that it isn't just for esoteric scientific applications - parallel computing also allows us to perform more mundane tasks such as video compression and sensor control in real time and with less risk of failure.www.ece.ucsb.edu/totem/
The triangle connecting ATM, network, and optical marks the hottest area in network research. On the bleeding edge, a team at Princeton University is putting together a local area network that will operate at 100 Gbps and faster. Today's electronic processors can't match those speeds, so the researchers are using optical processors for ATM routing. www.ee.princeton.edu/ooe
Medical imaging is one of the key applications driving the development of high-speed networks.
A project at the University of Hawaii at Manoa will enable real-time, 3-D medical images to be sent via satellite or fiber-optic networks. This would allow images to be sent from a remote disaster area in Hawaii to an Ohio supercomputing center for processing and enhancement, and then to a clinical specialist in Washington, DC - all in a matter of minutes.
