Hanover, New Hampshire

Tribeca is in a state of physical recovery tempered by emotional exhaustion. I'm spending more and more time away because I find it hard to lift myself out of the neighborhood's heightened sense of fate. Whenever I leave it takes a day or so for the concerns of ordinary life to seem real, but it's vital that they do.

This Summer I'm teaching and doing research part time up at Dartmouth with my long time friend and colleague Joe Rosen. Joe is best known as both as one of the world's most ambitious reconstructive plastic surgeons and as one of the most extravagant thinkers about how the human body might be modified in the future. There was a silly cover story about him on Harper's magazine a while back that dubbed him "Dr. Daedalus" because of his project to give people the option of having wings. We've worked together for almost two decades on surgical simulation, prosthetics, surgical instrumentation, and ways to use information systems to improve the health care system in the real world, right now, which turns out to be a harder problem than all of the above, because it involves lots of stubborn humans.

Joe and I work well together, but also clash over the same set of ideas that often put me at odds with other edge.org respondents. We're co-teaching a class on the future of information systems in medicine, and I think we might have actually succeeded in startling some of the almost perfectly serene students who have populated universities in the last few years with our argument.

Joe believes that Lamarckian evolution is coming back (it probably played a role in the earliest origin of life). He's a plastic surgeon and as far as he's concerned the genotype only serves to give him building blocks to remake the phenotype. People will invent themselves in the future and the phenotype will be unlatched from the genotype.

I wasn't about to let him get away with that! A funny response would be to claim that current fashion would have us believe that ideas are Darwinian (Memes!), so Darwin would still be guiding his hand in the operating room as future generations of kids pay to be turned into mythical beasts. But the more important response is based on the scale of computation. Joe's been working on wings for people for years and can't say how long it will take him to have a viable design. That's because it's actually a hard problem. The reason it took evolution a long time to make something like a hand is that it's hard and evolution was doing significant work that took significant time.

Mathematical equations are eternal and platonic, but computation always takes time, makes heat, is messy, and screws up repeatedly. And then there are the legacy, lock-in, and brittleness problems of software, which are even worse. It's tempting to imagine you can have the payoff of computation without having to pay the price. I'm always amazed that people are ready to imagine that we're about to be able to compute and simulate, and therefore design, anything we want just because computers will keep on getting bigger and faster. No! We have hard slogging to do! The point isn't whether the phenotype remains stuck to the genotype, but how much work it is to change a phenotype. If we think genes are inefficient at the job, we might just be fooling ourselves.

What we need now is to develop our intuitions about how hard biological-scale problems really are. The computers of thirty years from now will undoubtedly make human wing design easier, but not easy.

Of course, by the time we've argued this far, the notion that people will want wings has been sneakily and smoothly slipped into the set of assumptions held by the students. But will people in the future really want wings?

The polite and well spoken students of 2002 are often enhanced by tongue piercings and other ornamentations that represented rebellion fifteen years ago. At that time I had assigned a class of pierced students the problem of inventing what their children would have to do to shock them. That was a hated assignment! What is most surprising today is the lack of rebelliousness in the new crop. They would probably choose wings, but not in any provocative way.

Another little adventure in the last week was the premiere of a new movie by Steven Speilberg called Minority Report. Spielberg convened a collection of experts early in the movie's development to conceive of a world fifty years from now, in which the action would take place. We met in secret, and it was all very glamorous. I provided ideas for the computers, advertising, and communications technologies. Sending ideas through the Hollywood process and seeing the end result makes me sympathize with genes, which I can imagine being shocked or bemused from time to time as the creature they define from a great computational distance is revealed.

Some comments on gizmos in the movie: There's a grainy, wobbly tele immersive recording of the Tom Cruise character's ex-wife. The defects look a lot like tele-immersion demos from two years ago, because, I was told, a cleaner signal wouldn't tell the story. There's one technical mistake: The camera pans around to the side of the "autostereoscopic" display and the image still sticks out into space. An autostereo display lets you see 3D things as if they were floating in front of you, BUT no one can make a photon make a right angle turn in midair, and that's how they ended up designing the shot. I have to say the result works well cinematically.

The advertisements that sense Cruise's face as he walks by and place him into their designs are based on a real demo of machine vision techniques for finding and interpreting the faces of people in a room and incorporating them into virtual worlds. My initial thought was that Cruise would be trying to run away, but billboards would add him to their imagery as he passed, making it impossible to hide. The demo the scene is based on is from a little company called Eyematic (a spin-off from USC), based on research by Christoph von der Marsburg, Hartmut Neven, and others including me (I'm the Chief Scientist of the joint).

When I demoed Eyematic's face-grabbing software, I mentioned in a macabre aside that face matching to determine a person's identity might sometimes be a better idea than, say, iris matching, because criminals might be tempted to gouge out someone's eyes to fake out an iris-based system. (Not that it would actually work; living and severed eyes can be distinguished by machines. ) Well, you can see the result in the movie. Everything's based on iris-based IDs, and criminals merrily go about gouging out their own eyes, and those of anyone else who will pay for the resulting privacy. I had to close my own eyes during the eye-gouging episode. The severed eyes turned into fabulous props that convey comedy, suspense, and horror at the same time.

And then, most surprisingly to me, the iconic VR input device of the 1980s, the instrumented glove for handling virtual objects, makes a comeback.

The pattern that emerged was that devices that were already antiquated in 2002 proved to be more photogenic than the perfected devices I suggested for 2050.

Let's see, what else? There's music! It looks like I'll be working on two opera projects simultaneously this Summer. One is "Bastard the First", a collaboration with Terry Riley, and the other is "Mount Analog", based on the novel by Rene Daumal, working with a French artist named Philippe Parreno. I did some instrumental work for "Logic of Birds", a performance that Sussan Deyhim will give at Lincoln Center this Summer. The most fun was getting a contrabassoon to play in a Persian style.

And yes, I am working on my very overdue book, and someday soon I will shock everybody with a manuscript.


Postcards from: