Total Recall: How the E-Memory Revolution Will Change Everything (18 page)

Jim Gemmell was thinking about stories in 2005 when he asked me to spend a number of lunch breaks with the half-dozen others in the lab answering questions about my experience at Digital Equipment Corporation. We ended up with an hour of video stories and opinions never expressed elsewhere. I’m sure the viewer would get a different sense of me from seeing these videos than he would by reading what I have written about life at that time.

I believe oral histories are irreplaceable. While I love all the artifacts at the Computer History Museum, I’ve come to see the stories they collect in oral histories and public talks as the most important thing the museum does. Of course, memories fade over time, so for accuracy we want more than just oral histories. Still, inaccuracies and all, I want to capture people recounting stories.

For Total Recall to fulfill its full potential, people must be able to tell stories anytime, anyplace, any way they feel like it. This could be by sitting in front of video camera each night for a dear-diary entry, talking about some episode while driving, or typing in some thoughts about a recorded event.

AUTOMATIC SUMMARIZATION

With today’s technology, your e-memories would be a mixed blessing for your heirs. They would have the benefit of more knowledge about you, but it would come to them as an enormous, daunting mess. Your heirs may enjoy looking at random photos, or searching for e-mails containing the names of presidents in order to read some of your political perspectives, but they will likely miss the most important and interesting bits, and may be too intimidated to spend much time with your e-memories.

I felt this way about my own early scanned collection, and it was my frustrated eruption of “It’s just bits!” that galvanized Jim Gemmell into action and got us started on MyLifeBits. Thankfully, our prototyping work has assured me that things will get much better.

We will see the evolution of software that will reduce the chores involved in making one’s life bits worthwhile to others. It will help to develop tools to make storytelling and human arrangement easier. But fully automatic approaches are even more important. Some of this we get quite readily just by storing more information together. For example, your e-mails, calendar entries, Web-page visits, and digital photos all have time stamps that readily lend themselves to time line displays.

Time lines are a really compelling way to visualize your life, and software can help automatically produce digestible time lines. Our colleague Eric Horvitz and his research team have done some very promising work in predicting which events people will consider to be significant “memory landmarks.” This allows the best material to be put on a time line, and the less interesting material hidden away until it is asked for. Eric demonstrated the software to a reporter, starting with pictures of his wife and son:

“What’s cool—I love this feature—I can say, ‘Go to July Fourth,’ and it’s making guesses about the things I am likely to remember, to use memory landmarks, and it jumped right to this place,” he said. The screen showed several images—a small-town parade, and his wife and son among figures at a cookout, from July 4, 2005. Responding to his request, “the computer brought up its best guess.”

“It comes to understand your mind, how you organize your memories, by what you choose. It learns to become like you, to help you be a better you.”

Remember the Dublin City University project that finds novel SenseCam images out of thousands to highlight the interesting ones and spare you from the mundane? Automatic summarization has become an entire field of research. There are even sub-specialties, for example, summarizing just video. While we would like it best if a human creates a photo album, it is possible for the computer to do a pretty good job of automatic photo-album composition, choosing only interesting, high-quality photos. The nice thing about machine-composed albums or time lines is that, unlike physical albums, they keep the “outtakes”; with just a few mouse clicks you can retrieve the other shots that weren’t included in the album, allowing you to be as absorbed as you like with one particular event or topic.

With automatic summarization, posterity will be able to browse your e-memories starting from a manageable “birds-eye” view of a life, rather than just confronting an intimidating jungle of material.

IMMORTALIZING JIM GRAY

When Jim Gray went missing in 2007, I was not alone in wishing to immortalize him in the most rich and resilient way possible.

I am certain that Jim Gray’s name will be immortal at least in some ways. His name cannot be neglected in any history of computing as a winner of the Turing Award (often called the Nobel Prize of computer science). He is best known for his role in developing transaction processing, which we all use every time we withdraw cash from banking ATMs. In an effort to have his name be even better remembered, I helped establish the Jim Gray Endowed Chair in Computer Systems at UC Berkeley. I’d also like to see a building named after him. Jim’s astronomer friends have already identified an asteroid that will bear his name.

For a computer scientist like Jim, the most common way to gain an immortal name is to pass on ideas that are used by future generations. If you are lucky, some concept will be named after you. Moore’s Law is undoubtedly the best-known, predicting that transistor density in computer chips would double every two years, and explaining the meteoric rise of computing power. I hope that someday we will refer to the
Gray Data Cube
, the
Gray Transaction Processing Benchmarks
, the
Gray Five-Minute Rule,
and the
Gray Paradigm of Scientific Discovery
.

I’ll certainly be lobbying for such an immortal name for Jim. However, while it would be fitting for him to join the ranks of other such esteemed names, having an immortal name is a pretty superficial immortality. We may say Pythagoras’ name until the end of time when discussing geometry, but we will never know much about him, or even how he did his work. I want something better for Jim.

Someone’s work can be immortalized, as in the paintings of the great masters, buildings by a brilliant architect, or some notable equation. Going deeper, the
way
they worked may be immortalized: their techniques, their approaches, their professional relationships, and the stories of them at work. For instance, we know a fair bit about the work of Isaac Newton, including the story of him in his early twenties, going to the countryside to avoid an outbreak of the plague and, like any typical young man with too much time on his hands, whiling away his time—inventing calculus and discovering the law of gravitation.

Jim Gray’s Web site reveals a lot about him as a computer scientist. The extensive publications on the site reveal his drive for understanding through experiments and measurements. However, the site is missing those additional, critical stories that help us understand the man at work. Other sites, like the National Library of Medicine’s “leaders in biomedical research and public health Profiles in Science” Web site (
http://profiles.nlm.nih.gov
), provide a slightly more personal look. At this site, luminaries such as Francis Crick (who discovered the structure of DNA) have archives that include articles, bibliographies, books, brochures, certificates, drawings, exams, interviews, lectures, letters, notebooks, photos, and schedules. Still, we end up with only a fragmentary view of their lives.

Jim Gray’s family, friends, and colleagues are the sole repository of what he was really like: how Jim would slap your back with bubbling enthusiasm when he congratulated you, or how, when he thought your ideas were nuts, he would politely pronounce himself “puzzled” and furrow his brow. His Web site doesn’t tell you of his countless lunches on the least expensive sandwiches in San Francisco, despite his wealth. Then there are all the stories from his family, his sailing buddies, friends from college days, and others.

The story of Jim Gray is spread out on the computers he used, in his personal effects, and in the e-memories and bio-memories of those who knew him. Doing justice to his story means bringing them all together and presenting them in a comprehensible way.

MY MEMORIES OF JIM

After losing Jim, I naturally reminisced about my own relationship with him. A quick search in MyLifeBits turned up the following items that reference Jim in some fashion: 13,000 e-mails, 1,600 Web pages, 100 presentations, 289 photos, 600 documents, several videos, and a phone call. I don’t recall what first prompted Jim and me to get together, but MyLifeBits has a copy of the penciled calendar entry from 1994 for our first meeting. Other calendar entries include Jim taking me out on his sailboat—the same one he would later vanish in.

In 1994, Jim had just finished four years heading DEC’s San Francisco Lab on Market Street and had turned consultant. Since 1989, I had been a Silicon Valley angel investor and a consultant to Microsoft Research and others (since I didn’t spend much time consulting, some friends kidded me that
consultant
was a code word for “unemployed”). Our first meeting at my Los Altos home revealed our shared views on the importance of industry standards and an approach to increasing computing power via many cheap PCs working in concert together. We found that we both preferred small teams and esteemed building influential prototypes. It was the beginning of a stimulating collaboration and a heartfelt friendship.

After being an independent consultant for a while, Jim felt that he needed the confines of an organization, and he convinced me that I needed more structure too. He had been talking to Microsoft. We believed Microsoft was the place to be because of how we felt about standards and leverage, and moreover the respect and enjoyment of the community we would be part of. I jumped the gun and e-mailed the Redmond folks to hurry up and start a Microsoft Lab in San Francisco for Jim:

Sun Jan 08 15: 41: 55 1995

To: Rick Rashid ; Nathan Myhrvold ; . . . Dave Cutler

From: Gordon Bell

Subject: Approaches to Servers and Scalability . . . and an AD Lab here!

>Folks, Here’s how Jim Gray and I see the next decade or two: A Scalable Network and Platforms (SNAP) architecture is predicated on one set of standards: an ubiquitous ATM network
and PC-sized platforms.
SNAP allows upsizing i.e. building world-scale computers from a single platform architecture in a scalable fashion. SNAP will encourage further industry de-stratification. It eliminates the traditional computer price class distinctions (mainframes, minis, PCs) and goes a long way to eliminate the stratified business models of traditional computer suppliers. SNAP will cause a computer industry upheaval greater than the early 1990s client-server downsizing wave. That wave created a large UNIX market displacing IBM mainframes and proprietary minis. But the UNIX market is fragmented and small when compared to Compaq and NT. UNIX would have to consolidate around one or two dialects in order to get the volumes required to compete with NT. This seems improbable, so Microsoft’s NT is likely to become the dominant server standard for all hardware platforms, just as Windows garnered the desktop or client side.

Jim sent his own e-mail, pointing out that he had not “put me up” to writing mine, and enumerating the difficulties of operating a remote lab. However, he strongly validated the outlined vision. Microsoft liked the idea, and Jim’s Bay Area Research Center (BARC) opened in the summer of 1995 in San Francisco. I was honored and delighted to join the lab in August of that year. We hired Jim Gemmell to work with me that fall.

Though the BARC lab peaked at only around ten members, it had an impact beyond its numbers. Aerial imagery of the world was brought to the Internet by the BARC Terra Server, which led to the Microsoft Live Maps site and predated Google maps by five years. Later, Jim would turn the view up to the heavens, and work on the Sky Server project. His broad agenda got him involved in such far-flung projects as the “land speed record” for network transmission and fail-safe databases. Meanwhile, Gemmell and I were working on telepresence: putting a conference on the Web, playing with altering someone’s gaze direction in video, and shipping new network protocols in Microsoft operating systems. Later, of course, we got into MyLifeBits.

A memorable event was in May 1997, when Jim gave an on-stage demo with Bill Gates, using more than a hundred PCs to achieve one billion transactions per day. I also recall Jim’s glee on April Fool’s Day 2005, when he had just finished measuring a half-billion transactions per day using his relatively old laptop. He wrote a report observing that a common PC could execute eighty times “more than one of the largest U.S. bank ’s 1970s traffic—it approximates the total U.S. 1970s financial transaction volume. Very modest modern computers can easily solve yesterday’s problems.” The data and report illustrate Gray’s fondness for understanding through constant building and experimentation.

Through various paths, Jim infected me with the importance of data. It’s “all about the data,” he would say. In one of our more playful times, while discussing how to get the concern for data into the national computing resource allocation agenda, we bumped into John Markoff, a friend and columnist at
The New York Times
who also had an office in our building. We made the case that the national computing agenda was missing the point by just thinking about computation speed. John took our picture in the lab on Friday and wrote an article that appeared in the
Times
on Sunday. Our compute-centric friends in Washington were not especially happy, but they slowly came around to our view, and after many years the situation is gradually changing from models and simulation to real world, data-based, data-intensive science.