A few of the things on Gordon Bell's mind: From ancestral photos to his MIT diploma to his robot driver's license, it's all in there. And we do mean all.
One of Bell's Microsoft allies is also investigating whether artificial intelligence could be used to find hidden patterns in memory. One day last summer, I visited Eric Horvitz, an expert in machine intelligence at the Research Labs, to see his "Lifebrowser." The Lifebrowser's goal is to automatically identify the most significant events in your life, so that when you scroll back through your history, it shows you only the most important highlights.
The software starts by having you "train" it by rating different things on your computer as significant or negligible. But then you turn Lifebrowser loose, and things get really interesting. It begins to observe your daily behavior and rank your documents and calendar events. Horvitz has given the software to members of his team at Microsoft, and Lifebrowser has already discovered several intriguing things about how we work. Unusual events--"atypia"--tend to be the ones that people most want to remember. Intuitively, this makes sense: Sudden, unexpected news, good or bad, is highly meaningful; a large cluster of pictures from the same day usually means you saw something odd or important--pluck one of them out, and it's likely to be a rich memory. (In contrast, regular meetings are virtually never remarkable. "No one ever needs to remember what happened at the regular Monday staff meeting," Horvitz notes dryly.)
He pulls me over to his computer screen to see his own personal Lifebrowser. At first, it just looks like a regular calendar, with months stretching back years.But when he zooms in on October 2004, I can see that Lifebrowser has carefully picked only a few items to display on each day: A meeting at DARPA, the military-research agency. Pictures of Horvitz's family visit to Whidbey Island. An email announcing a surprise visit from his old college friend.
"I would never have thought about this stuff myself," Horvitz says, scanning the calendar. "But as soon as I see it, I'm like, 'Oh, right--that was important!'" Lifebrowser, in essence, shapes the seemingly random flow of our lives and reminds us of what we ought to be focusing on.
For me, though, the real holy grail would be a system that doesn't merely remind me of what I already know--or once knew, even. It would be one that actually conceives of new ideas.
Few computer scientists are claiming to have figured that one out. But a hint of what may be coming has already arrived from Devon Technologies, a company that has created a tool called DEVONthink. It works like this: As you work on a project, you feed DEVONthink copies of any emails, memos, documents, PDF files, or Web pages you find interesting. The software performs a complex analysis of all that information, trying to find documents that are statistically similar to one another in meaning. This is much more sophisticated than a simple word search, though: DEVONthink can learn that the words "automobile repair" and "car fixing" are synonymous, for example, even though they use completely different terms. This allows the software to spot some astonishingly subtle connections.
Steven Johnson, a writer in New York, discovered that DEVONthink's smarts could literally inspire him to come up with profitable new ideas. He had been using the software for several years, putting all his research into it. While writing his latest book last year on the London cholera outbreak of 1854, he'd plug paragraphs of his draft into DEVONthink to see if anything else on his hard drive was related. One day, he inserted a passage on the London sewer system that used the word "sewage" a lot. DEVONthink unexpectedly unearthed a quote that discussed how vertebrates originally evolved their bones--by reusing the calcium waste generated by their metabolism.
It was a weird connection: comparing the waste systems of a city to those of an animal. But a lightbulb went off in Johnson's head, and he began dreaming up a new chapter for his book, contrasting the ways that cities and the human body find productive uses for their by-products.
"Now, strictly speaking, who is responsible for that initial idea?" he wondered, when he described the experience in an article. "Was it me or the software?" The idea, he argues, was a synthesis of two minds--his real brain and his virtual one. "Two very different kinds of intelligence playing off each other," he says, "one carbon-based, the other silicon."
In spring 2004, Gemmell lost a chunk of his memory. The Microsoft senior researcher had built his own personal MyLifeBits database, filling it, like Bell, with oodles of his email, Web surfing, and pictures. But one day, Gemmell's hard drive crashed, and he hadn't backed up in four months. When he got his MyLifeBits back up and running, the hole that had been punched in his memories was palpable, even painful. He'd be working at a project and vaguely remember some Web site or document that was important, then begin drilling down to find it--only to discover that it was part of the missing period.
Recent Comments | 9 Total
September 25, 2009 at 12:17am by Christopher Jeschke
Nice picture!
--
Photo Blog