Total Recall: How the E-Memory Revolution Will Change Everything (19 page)

DIGITAL IMMORTALITY

I’ve just told some stories about Jim Gray, and shared an e-mail artifact. But if my publisher would let me, I’d probably include a photo; and in the e-book, some audio and video too. If Jim had lifelogged, there would be a wealth of e-memories to peruse.

A complete lifelog is an awesome prospect. Passing on your e-memories might be seen as a sort of one-way immortality, allowing you to communicate to the future. But Jim Gray and I used to speculate about the possibility of something really worth calling digital immortality, where a digital version of yourself lives on and interacts with posterity. What if, a hundred years from now, your heirs could ask you questions and you could answer?

You might be familiar with the word
avatar,
meaning an image or computer-generated character that represents someone in cyberspace. An avatar might be a picture of you that appears while you instant message with someone. Or, it may be a three-dimensional character in a virtual-reality program like Second Life. Someday, those who lifelog will be able to create avatars that do an amazing job of impersonating them.

Imagine if you could have a conversation with an avatar of your great-grandfather, seeing his face, hearing his voice, and having your questions answered in your great-grandfather’s unique turn of phrase. It seems clear that at least some aspects of your person can be expressed as a program that actually interacts with future generations. Lifelogging is the key.

Your lifelog will have all the details of how you sound, the phrases you employ, questions you have answered, and facts about your life. It will contain recordings of you under stress and relaxed, pleased and annoyed, in triumph and in defeat. It will know your favorite quips and mottos. It will have pictures of you that can be overlaid on a three-dimensional model. If someone should want to impersonate you, your lifelog will be the ultimate gold mine.

There has already been some progress making avatars that answer as the individual would have. Carnegie Mellon University researchers have created a program that lets you interview Albert Einstein. I’ve invested in a company called MyCyberTwin, whose software trains an instant messenger (IM) avatar how to answer as you would. You just give it content from your own IM conversations, e-mail, blog, et cetera, and it learns how to imitate you—it becomes your “cyber twin.”

MyCyberTwin analyzed transcripts of the television show
The Simpsons
to create a cyber twin for Bart Simpson, who had the following chat with me:

You:
What’s up?

Bart:
Well, I was interested in getting the fire hydrants from the school to power my downhill racer—wanna help me?

You:
No. How is Lisa?

Bart:
She can be so annoying, but she is still pretty cool. Do you like my Mom’s hair?

You:
Blue hair is pretty cool too. What do you think of your dad?

Bart:
My dad, Homer . . . he got a chess set once, then packed it away. When I asked him why, he said he is saving it for a brainy day.

Have you been married to my aunt Selma?

You:
No. How old are you?

Bart:
I’m ten, man. Best age in the world: You can do what you want without getting in too much trouble.

You:
What do you want to do when you grow up?

Bart:
It’s always important to consider the consequences of anything you do.

Are you a friend of Principal Skinner, by any chance?

You:
Do you have a pet?

Bart:
Who needs pets when I have Homer? Actually, I do have Santa’s Little Helper too. He’s a dog. What pets do you have?

MyCyberTwin specializes in creating chat avatars to perform customer service, for example, answering questions for bank customers. Their avatars have gotten better scores than humans in accuracy, sales performance, and customer satisfaction. Now the MyCyberTwin folks are intrigued by the idea of taking my own e-memories as input—there is enough of what I have said in e-mail, letters, chat, papers, and so forth, that one ought to be able to construct a pretty realistic Gordon Bell cyber twin.

Alan Turing, a founding father of computer science, proposed the Turing test for determining a machine’s capability to demonstrate intelligence: A human judge has a conversation with a human and a machine, each of which tries to appear human. If the judge can’t tell which one is human, then the machine has passed the test. Turing proposed typewritten exchanges; we can update that to computer chat without changing the essence of the test. Thus, we can have a cyber-twin test: You chat with someone and his cyber twin. If you can’t tell the difference, then the machine passes the test. Note that the cyber twin could have a much better memory than the human did; it should be taught to forget in a similar way to the human for real simulation. But it should remember when we really want the answer! As I write, there is a fair bit of work remaining before any cyber twin could pass the test, but substantial progress seems likely.

I see four steps in the progression of digital immortality. First is digitizing the legacy media one has. Second is supplementing one’s e-memories with new digital sources. The third is two-way immortality—the ability to actually interact with an avatar that responds just like you would. The fourth is an avatar that learns and changes over time just as you
would have.

Having an avatar that actually learns and grows over time is a much more speculative idea. Who’s to say that it is working correctly? If we could predict how someone would behave after death, then we could predict people during life—an idea that sounds far-fetched. Perhaps a more realistic goal would be to shoot not for growth—as in change—but just accumulated knowledge, so the machine would recall when it last communicated with you and what was said: “Hi Gordon. I talked to you yesterday. You told me about your vacation.”

While I believe the fourth step will remain science fiction, an avatar passing the cyber-twin test is not. A lot of sci-fi and artificial intelligence discussion is about true machine intelligence, where programs actually learn and grow. Personally, the more I learn about machine intelligence, the more I am impressed with the learning ability of the average two-year-old. True machine intelligence remains elusive. Computers can beat human chess masters now, but the computer’s greatest advantage is its ability to enumerate each and every possible move and outcome; most of us would call that tediously mindless, not intelligent. Similarly, in an attempt to answer factual questions like “What year was Abra ham Lincoln assassinated?” brute-force approaches that just scan many encyclopedias and newspaper sources looking for common words often do as well as or better than programs that attempt to parse and comprehend the same texts. I am confident about avatars passing the cyber-twin test because I see that lifelogs will contain enough information to support such brute-force approaches. No sci-fi, truly intelligent machines are required.

By carefully mining your lifelog, we should be able to ask your e-memories questions and hear your answers. We can change the game from a search to a discussion.

Humans have a natural propensity for recording life. Just look at all the people walking around with cameras and video cams. You’d be hard-pressed to find a home without photo albums, home movies, scrapbooks, and mementos. The one thing many people would be sure to rescue from the flames of a burning home would be their photo albums. We love to reminisce, and if you think of all the photos and home movies taken, it seems we enjoy
enhanced reminiscence
: not just remembering but also hearing and seeing recordings or artifacts from the past. A few of us go beyond just confining ourselves to recordings and objects, and actually edit movies, or create scrapbooks with captions and artistic layout. Some even take classes from companies like Creative Memories to learn to do it better. The rest of us envy them the time and talent to produce such compelling stories.

Your e-memories will prepare you for the digital afterlife. Already, for a fee, Web sites like
www.legacy.com
and
www.forevernetwork.com
offer to store letters, essays, photos, videos, and stories to pass on to future generations.
Famento.com
helps people share family stories, building e-memorials to loved ones. These sites are the digital equivalents of cemeteries and libraries.

Imagine opening up an old tomb and finding it full of historic artifacts. That would be interesting for a while, but just think of how much more interesting the artifacts would be if they came with their own curator, ready to help guide you through and explain them. That’s what I expect Total Recall software to do for our e-memories, with automatic travelogues, automatic summarization, and cyber twins. Today, as I struggle with setting up Jim Gray’s digital legacy, I know we’ve got a long way to go. But I also know that Jim would have insisted on seeing this as a challenge—another chance to do and apply science by creating understanding and something of value.

PART THREE

CHAPTER 8

LIVING THROUGH THE REVOLUTION

New technologies have always forced mankind to adapt to new realities, from iron tools to mobile phones. The changes worked on our societies by powered machinery were so radical that we refer to the industrial “revolution.” We are now beginning the Total Recall revolution.

I’m a technologist, not a Luddite, so I’ll leave abstract discussions about whether we should turn back the clock to others. Total Recall is inevitable regardless of such discussions. However, as a realist I also know that we must come to grips with the implications of our technology. Some implications I see as “bugs” to be fixed. For example, an unresolved bug of the industrial revolution is pollution. Other implications are simply changes that we must adapt to, such as modern transportation implying that one can commute to work and get fresh fruit from another continent.

This chapter is about the changes generated by Total Recall, both the bugs to be fixed, and the adaptations that will be required. These changes are mostly about what happens to our e-memories once we have them. Could we lose them? Could they fall into the wrong hands? What is their proper use? Different cultures may come up with different answers. Technology will yield unintended negative consequences and pleasant surprises too. If the answers aren’t all clear, most of the questions are.

DATA LOSS AND DECAY

Right now we face data decay and loss. Data often only exists in one place, so a crash of the host device means permanent loss of the data. Files formats may become unreadable over time. We need improved data longevity.

One morning in the fall of 2008 my notebook computer wouldn’t start up. I was in Australia and it was all I had. I broke out in a sweat. Questions fired off in my mind: Is the hard drive shot? How much isn’t backed up? How will I get my e-memory back in action? I decided the biggest problems would be a presentation and some articles if the disk was bad, because they had no backup. I also faced the hassle of paying bills without Microsoft Money, which I’ve come to count on.

Fortunately I had backed everything up onto my assistant’s machine in San Francisco two weeks before getting on the plane. I could download my e-memory as of two weeks ago from her, or I could just use it straight off her machine when I was on the Internet. My most recent e-mail was intact because it was on our corporate e-mail server.

This episode reminded me that I need to be prepared to lose whatever is not backed up. I had just finished authoring a presentation and was crazy not to have copied it to a USB thumb drive.

I believe the chief obstacle to data longevity is low expectations. For too long, too many of us have been content to see the data in our PCs, PDAs, and cell phones as transient. We shrug at losing the phone numbers from our cell phones, or not being able to do anything with files on an old floppy disk. This tolerance for data decay was natural when personal computers were new and few people had any experience with electronic storage. Fortunately, a couple of decades into widespread computer usage, we are learning better. I’m encouraged by signs of this trend, for example, seeing that Dell notebooks ship with Dell DataSafe Online software installed to perform backups to their site.

Protecting against outright data loss involves two techniques: replication and backup.
Replication
means that a copy is made of every bit of data you own. The more copies, the better. It is best to make copies that are located far away from each other, so that a hurricane, earthquake, or fire doesn’t destroy all of the copies at once. Such geographic replication has been commonly employed by Fortune 500 companies for many years; a bank cannot tolerate even the thought of losing all its account balances.

Backup
is a little different than replication. A replica is important, but what if you accidentally change an important file? The next day you look at the file and realize you’ve wiped out some valuable information. You can’t turn to a replica, because it has faithfully copied your destructive changes. A backup is a snapshot of your data at a given moment, to cover you in the event that you need to get back to an older version.