After departing the D conference and fighting morning traffic on Interstate 5, I settled into a quite different conference experience. D was held at the Four Seasons Aviara, a plush and modern hotel sitting on a hillside five miles inland from the Pacific. The conference itself was held in one of those faceless hotel ballrooms, with a fabricated stage complete with a 6-foot letter D.
Interview subjects sat in hefty red leather armchairs that looked almost exactly like dentists chairs. Attendees sat not at desks, but in plush black leather armchairs or comfy office-style swivel chairs. The D audience—made up as much of journalists and PR people as experts and CEOs—was a rambunctious, querulous lot.
FIRE (Future in Review), in contrast, is held at the aging jewel of the San Diego coast, the Hotel Del Coronado. The semicircular ballroom looks out over the Pacific Ocean, and the stage looks dressed more for a production of "My Fair Lady" than for a tech conference. Four simple chairs sit around a coffee table, with a battered podium offset in the corner.
The audience is decidedly low-key too. More international, and more cerebral, its mostly investors, researchers and chief technology officer-level technologists. The focus is different too. Instead of getting an update from Gates, Jobs and their ilk, the panels are focused on the future—although Gates nemesis Szulik, CEO of Red Hat, was the key afternoon interview. The D conference was like attending a college pep rally, chock full of overachieving jocks and circle-pin wearing coeds. FIRE, on the other hand, feels like the science club: geeks and wonks.
How did the geeks do? The best part of the day was the last, where the audience split into four groups. I trundled off to the basement of the hotel to sit in on a CTO roundtable, featuring Fred Weber of AMD, Rick Rashid of Microsoft and Dick Lampman of HP. The conversation started by focusing on things that are on the threshold today, but were virtually inconceivable five years ago.
Rashid led off with two topics that will soon change everything.
"One is human-scale storage," Rashid said, "to keep all the relevant information a person might generate in a lifetime with them wherever they go."
Soon terabyte hard drives will be common, he went on; hard drives big enough to store every conversation youve ever had in your life, or to take a picture every minute, even while youre asleep. He described a research project in Microsofts Cambridge, England facility that is essentially a "black box for humans." Chock-full of sensors, accelerometers and motion detectors, it takes a picture every time something changes. Its being considered as a way of helping patients with non-severe memory loss, to let them keep and summarize memories at the end of the day. London police are even interested in using it to help solve crimes.
Regarding Rashids second imminent breakthrough, he said that we are almost to "the point at which LCD technology will be cheaper per square inch than whiteboards." And when that happens, "almost any surface becomes an input and output device" and, with small cheap projectors, "any table, any wall, any surface [becomes] an input-output device."
"Its not that far away," Rashid promises, "and these new kinds of applications change the way you think about computing."
AMDs CTO, Fred Weber, expanded on Rashids vision by describing the simplicity of an LCD. "Think about a mirror: Its just a flat piece of glass with some chemicals on the back. Whats an LCD? A flat piece of glass with some chemicals on the back." He went on to predict that soon LCDs would be as ubiquitous as mirrors.
Weber also pointed to the imminent rise of multiprocessing as another inflection point. "Weve crossed a threshold where multiprocessing computers will be ubiquitous," he said. And that, according to Weber, will result in "real innovation, because [developers] will be thinking of algorithms differently, thinking of them in a parallel way." He predicted that some amazing killer apps will soon emerge from having multicore processors on every desk.
Weber also said he was excited about the potential for more natural human/machine interaction. "We are at a threshold where language processing and the ability to have somewhat intelligent interactions with a machine are near. Its a great sink for compute power in the future. These devices [computers] are becoming unbelievably capable, but difficult to use. Its not because the developers are idiots, its because humans are difficult to use," he said.
Rashid agreed, and described an interesting Microsoft research project that uses statistical techniques to derive answers from Web searches. According to Rashid it really works, so if you ask it when Lincoln was shot, it will come back with the right answer, and a different answer if you ask when Lincoln died (as the two dates are not the same). The program does this by statistically evaluating all the answers to the question that it receives via a Web search, and then picking the most likely response.
"You can ask it some really crazy questions and it will give you funny answers," Rashid went on. Put in "Masked Man" and it comes back with "Lone Ranger." Ask it "What is the meaning of life?" and it will deliver "42," but the answer ranked number one is "Questions."
The research has broader implications, according to Rashid. "A lot of the problems we tried to solve by rules and structures are fundamentally about data. Collect enough data and analyze it, and we suddenly see whats going on. Theres a lot of work now around statistical analysis. All of a sudden we have that much more data to process," he said. Weber agreed: "Always remember that exponentials are magic. You are getting such rapid growth and nonlinear growth."
Weber also tied the three trends together, saying, "Weve crossed a threshold where life storage, natural language processing, and ubiquitous input-output are much more likely to happen in the measurable future. Five years ago you couldnt have said that."
The conversation continued, mostly focused on security. One last fascinating factoid: "It takes very little data to uniquely identify a human in the United States," Rashid said, quoting a recent study. 83 percent of the U.S. population can be identified with just three things: date of birth, sex and zip code.