Smarter Than You Think: How Technology is Changing Our Minds for the Better. Clive Thompson
have superb records, but they don’t search them unless, while using their own brains, they realize there’s something to look for. And of course, our organic brains are riddled with memory flaws. Bell’s lifelog could well contain the details of a great business idea he had in 1992; but if he’s forgotten he ever had that idea, he’s unlikely to search for it. It remains as remote and unused as if he’d never recorded it at all.
The real promise of artificial memory isn’t its use as a passive storage device, like a pen-and-paper diary. Instead, future lifelogs are liable to be active—trying to remember things for us. Lifelogs will be far more useful when they harness what computers are uniquely good at: brute-force pattern finding. They can help us make sense of our archives by finding connections and reminding us of what we’ve forgotten. Like the hybrid chess-playing centaurs, the solution is to let the computers do what they do best while letting humans do what they do best.
Bradley Rhodes has had a taste of what that feels like. While a student at MIT, he developed the Remembrance Agent, a piece of software that performed one simple task. The agent would observe what he was typing—e-mails, notes, an essay, whatever. It would take the words he wrote and quietly scour through years of archived e-mails and documents to see if anything he’d written in the past was similar in content to what he was writing about now. Then it would offer up snippets in the corner of the screen—close enough for Rhodes to glance at.
Sometimes the suggestions were off topic and irrelevant, and Rhodes would ignore them. But frequently the agent would find something useful—a document Rhodes had written but forgotten about. For example, he’d find himself typing an e-mail to a friend, asking how to work the campus printer, when the agent would show him that he already had a document that contained the answer. Another time, Rhodes—an organizer for MIT’s ballroom dance club—got an e-mail from a club member asking when the next event was taking place. Rhodes was busy with schoolwork and tempted to blow him off, but the agent pointed out that the club member had asked the same question a month earlier, and Rhodes hadn’t answered then either.
“I realized I had to switch gears and apologize and go, ‘Sorry for not getting back to you,’” he tells me. The agent wound up saving him from precisely the same spaced-out forgetfulness that causes us so many problems, interpersonal and intellectual, in everyday life. “It keeps you from looking stupid,” he adds. “You discover things even you didn’t know you knew.” Fellow students started pestering him for trivia. “They’d say, ‘Hey Brad, I know you’ve got this augmented brain, can you answer this?’”
In essence, Rhodes’s agent took advantage of computers’ sheer tirelessness. Rhodes, like most of us, isn’t going to bother running a search on everything he has ever typed on the off chance that it might bring up something useful. While machines have no problem doing this sort of dumb task, they won’t know if they’ve found anything useful; it’s up to us, with our uniquely human ability to recognize useful information, to make that decision. Rhodes neatly hybridized the human skill at creating meaning with the computer’s skill at making connections.
Granted, this sort of system can easily become too complicated for its own good. Microsoft is still living down its disastrous introduction of Clippy, a ghastly piece of artificial intelligence—I’m using that term very loosely—that would observe people’s behavior as they worked on a document and try to bust in, offering “advice” that tended to be spectacularly useless.
The way machines will become integrated into our remembering is likely to be in smaller, less intrusive bursts. In fact, when it comes to finding meaning in our digital memories, less may be more. Jonathan Wegener, a young computer designer who lives in Brooklyn,17 recently became interested in the extensive data trails that he and his friends were leaving in everyday life: everything from Facebook status updates to text messages to blog posts and check-ins at local bars using services like Foursquare. The check-ins struck him as particularly interesting. They were geographic; if you picked a day and mapped your check-ins, you’d see a version of yourself moving around the city. It reminded him of a trope from the video games he’d played as a kid: “racing your ghost.” In games like Mario Kart, if you had no one to play with, you could record yourself going as fast as you could around a track, then compete against the “ghost” of your former self.
Wegener thought it would be fun to do the same thing with check-ins—show people what they’d been doing on a day in their past. In one hectic weekend of programming, he created a service playfully called FoursquareAnd7YearsAgo. Each day, the service logged into your Foursquare account, found your check-ins from one year back (as well as any “shout” status statements you made), and e-mailed a summary to you. Users quickly found the daily e-mail would stimulate powerful, unexpected bouts of reminiscence. I spent an afternoon talking to Daniel Giovanni, a young social-media specialist in Jakarta who’d become a mesmerized user of FoursquareAnd7YearsAgo. The day we spoke was the one-year anniversary of his thesis defense, and as he looked at the list of check-ins, the memories flooded back: at 7:42 a.m. he showed up on campus to set up (with music from Transformers 2 pounding in his head, as he’d noted in a shout); at 12:42 p.m., after getting an A, he exuberantly left the building and hit a movie theater to celebrate with friends. Giovanni hadn’t thought about that day in a long while, but now that the tool had cued him, he recalled it vividly. A year is, of course, a natural memorial moment; and if you’re given an accurate cue to help reflect on a day, you’re more likely to accurately re-remember it again in the future. “It’s like this helps you reshape the memories of your life,” he told me.
What charmed me is how such a crude signal—the mere mention of a location—could prompt so many memories: geolocation as a Proustian cookie. Again, left to our own devices, we’re unlikely to bother to check year-old digital detritus. But computer code has no problem following routines. It’s good at cueing memories, tickling them to recall more often and more deeply than we’d normally bother. Wegener found that people using his tool quickly formed new, creative habits around the service: They began posting more shouts—pithy, one-sentence descriptions of what they were doing—to their check-ins, since they knew that in a year, these would provide an extra bit of detail to help them remember that day. In essence, they were shouting out to their future selves, writing notes into a diary that would slyly present itself, one year hence, to be read. Wegener renamed his tool Timehop and gradually added more and more forms of memories: Now it shows you pictures and status updates from a year ago, too.
Given the pattern-finding nature of computers, one can imagine increasingly sophisticated ways that our tools could automatically reconfigure and re-present our lives to us. Eric Horvitz, a Microsoft artificial intelligence researcher, has experimented with a prototype named Lifebrowser, which scours through his massive digital files to try to spot significant life events. First, you tell it which e-mails, pictures, or events in your calendar were particularly vivid; as it learns those patterns, it tries to predict what memories you’d consider to be important landmarks. Horvitz has found that “atypia”—unusual events that don’t repeat—tend to be more significant, which makes sense: “No one ever needs to remember what happened at the Monday staff meeting,” he jokes when I drop by his office in Seattle to see the system at work. Lifebrowser might also detect that when you’ve taken a lot of photos of the same thing, you were trying particularly hard to capture something important, so it’ll select one representative image as important. At his desk, he shows me Lifebrowser in action. He zooms in to a single month from the previous year, and it offers up a small handful of curated events for each day: a meeting at the government’s elite DARPA high-tech research department, a family visit to Whidbey Island, an e-mail from a friend announcing a surprise visit. “I would never have thought about this stuff myself, but as soon as I see it, I go, ‘Oh, right—this was important,’” Horvitz says. The real power of digital memories will be to trigger our human ones.
In 1942, Borges published another story, about a man with perfect memory. In “Funes, the Memorious,” the narrator encounters a nineteen-year-old boy who, after a horse-riding accident, discovers that he has been endowed with perfect recall. He performs astonishing acts of memory, such as reciting huge swathes of the ancient Roman text Historia Naturalis and describing the precise shape of a set of clouds he saw