Question 1: Tell us about your background. When did you first hear of the concept of molecular nanotechnology? How long have you been doing nanotechnology research?

I first heard of Engines of Creation in the mid-eighties while I was a staff researcher at Rutgers working on AI and computer architecture. As a long-time SF fan, I was long since acquainted with vaguer concepts like "nanoelectronics" and manufacturing using atomic precision.

I had already tried to form a company to build self-replicating robots, along with the rest of the robotics fad of the early 80's, and I was
well acquainted the concepts on that side, having studied von Neumann's work.

In the late 80's I founded the sci.nanotech newsgroup, and moderated it for 12 years.  I began doing some nano-related work in the early
90's, e.g. reversible computing, and left Rutgers for IMM in the late 90's.

Question 2: Have you come across any specific technical criticisms of the concept of molecular assemblers that you find valid?

No.  Of course, it's difficult to give a technical criticism of a machine without even a design of the machine to go on, so that's not a particularly useful statement.  Most of the critics have had adopted the position "you haven't designed it yet so it must be impossible," which isn't particularly useful either.

Most critiques, like Smalley's "fat/sticky fingers problems", are based on a misunderstanding of what an assembler would really be like.  In particular, no one envisions picking up atoms as if they were golf balls.  We're just going to use the same kinds of reactions a chemist would use, except that we're going to control the positions of the reagents so the reaction can only take place where we want it to.

Question 3: Some nanotechnology enthusiasts are concerned that the articles by Whitesides and Smalley in the August 2001 edition of Scientific American will lead many to conclude that the concept of molecular nanotechnology is impossible. Are you concerned by the public reaction to Scientific American's issue on nanotechnology?

Not really.  SciAm has a long-standing hostility to Drexler not reflected in the general literature.  For specifics on the Aug. 2001 issue, have a look at http://www.imm.org/SciAmDebate2/ ... note that it's the second Foresight/SciAm debate!

Note that one of the reasons that scientists pooh-pooh assemblers and the like is to protect themselves from Bill-Joy-like Frankenphobia. There are many out there who would like to stop all nanotechnology research out of fear, and there's a strong incentive to say "Oh, don't worry about that, it can't happen, so you can keep giving me my research funds."

Question 4: Lyle Burkhead currently runs a site, (geniebusters.org), in which he argues that molecular nanotechnology will never be used to create bulk commodities. The basic crux of the site's argument is that a machine containing molecular assemblers will never be built because the complexity needed to design most objects (such as computer chips) requires layer upon layer of expertise and specialized skill. Each molecular factory would need to be so highly programmed that a whole industry would be needed to design, support and oversee the operation. If a whole industry is needed, he argues, then we would simply have the basic production system we have today shrunk to the nanoscale. He writes that "Designing, building, and retooling complex apparatuses won't be easy or free in the future any more than it is now". He further states that:

"Making things with atomic positioners will be at least as expensive as making them with biotechnology or bulk technology... atomic positioners will only be used to make things that could not be made in any other way."

What is your response to his criticisms?

Burkhead's first main point is that nanomanufactured stuff will be as hard to design and specify as current stuff is, if not harder.  That is quite right -- but he misses the point that the end user doesn't actually design and specify stuff, he/she just selects it.  The user of a synthesizer box might have an interface that looks like shopping on the internet.

His next point is about economics.  He makes some points that are quite valid, especially in the short run.  However you'll note that he himself has had second thoughts about his conclusions and amended some of his pages.

Here's a model for the economics of nanotechnology:  When I was a kid, you bought film for your camera, took pictures, wrapped up the exposed film, and mailed it to Rochester.  In the big central plant it was developed, and you got your pictures back a couple of weeks later. The big central factory cost many millions to build.

Next, in the past couple of decades, they built developing machines that allowed local businesses to appear, for capital costs in the hundred thousands, which did same-day developing with much greater convenience.

Now, of course, with a digital camera and color printer you can do it all yourself.  Convenience is maximized and capabilities are enormously expanded.  The cost (in time and money) of taking, making, saving, and sharing pictures is tiny on a per-picture basis, compared with a generation ago.

With molecular manufacturing, the same kind of thing will happen for 3-dimensional goods.

Burkhead then goes on to argue against AI making any difference, but ultimately criticizes his own argument and allows that computer-assisted humans will be much more capable in the future than they are now. That's the only part of the point that concerns us here, as far as the economics of designing nanosystems is concerned.

Question 5: Tell us about Utility Fog. How did you come across the concept, and how extensively could such a technology be used?

I invented Utility Fog in a typically serendipitous way.  Virtually everyone working with nanotechnology has had ideas for a polymorphic material to make objects out of.  But I was driving in to work one day and became conscious of my seat belt. I began wondering how good a seat belt you could make with nanotechnology. One of my pet peeves about safe cars is the they're built to collapse in an accident; the crumpling of the structure gives you a longer deceleration path, which is what makes it safe.

Suppose your vehicle looked more like a living room inside, with lots of space around you.  It would be a lot more comfortable than a conventional car, and there would be room to decelerate without trashing the vehicle. Next thought: the cases they ship delicate equipment in, with form-fitting foam interiors.  Suppose you could do a similar thing as a seat belt, in such a way that it didn't appear to be there when it wasn't needed. Once the basic notion was there, it remained only to figure out how to implement it.

Utility Fog consists of a mass of tiny robots.  Unlike water fog, they do not  float in the air but form a lattice by holding hands in 12 directions (corresponding to the struts in an octet truss).  Each robot has a body that is fairly small compared to its armspread, and the arms are relatively thin.  Each arm is telescoping, an action driven by a relatively powerful motor, and can be waved back and forth (2 more degrees of freedom) by relatively weak motors.

The material properties of this mass depend on the programming of the robots.  The geometry is such that stresses in the material all appear as longitudinal forces along the arms.  Each Foglet can sense the force along each arm, and do something depending on the magnitude and relation of those forces.  If the program says, extend when the force is trying to  stretch, retract when it is trying to compress, you have a soft material. If it says, resist any change up to a certain force, then let go, you have a hard but brittle material.

If the programming says, maintain a constant total among the extension of all arms, but otherwise do whatever the forces would indicate; and when a particular arm gets to the end of its envelope, let go, and look for another arm coming into reach to grab; you have a liquid. If you allow the sum of the arm extensions to vary with the sum of the forces on the arms, you have something that approximates a gas within a certain pressure range.  Note that because the Foglets can use their own power to move or resist moving, the apparent density and viscosity of the fluid can anything from molasses to near vacuum.

Now you can begin to get cute.  Run a distributed program that at a specified time, changes a certain volume from running water to running wood.  A solid object would seem to appear in the midst of fluid. It can just as easily disappear.  Now fill your entire house with the stuff, running air in background mode.  Have an operating system that has a library of programs for simulating any object you may care to; by giving the proper command you can cause any object to appear anywhere at any time.  You could carry a remote control, which might happen to be shaped like a wand with a star on the end...

Question 6: How would the individual foglets be able to communicate and coordinate with each other? The task of assigning each one of the quadrillions of foglets a precise place seems incredibly difficult.

Well, it is incredibly difficult.  Frankly, the capabilities of Fog will depend on the capabilities of software and other control technologies.  However, there are relatively simple methods, involving distributed physical simulations at large scales and "traffic rules" at small ones that should work for the gross physical motions involved with simulation of macroscale objects.

Wave your hand in the air.  How did the air molecules know to get out of the way, and to fill in the space behind?  Pressure waves, moving like ripples on a pond but at the speed of sound, caused differences in pressure in the air all over the room, and the air molecules moved, coordinated by bumping into each other, in response to the derivative of the pressure field.  There are intermediate levels of description involving currents and turbulence.  But each molecule doesn't know any of this; it only bounces off its neighbors.

Question 7: Approximately when do you believe that the first molecular assemblers will be built?

Between 2010 and 2020.

Question 8: Writers such as Ray Kurzweil claim that the pace of technological advance is accelerating. Critics of Kurzweil claim that exponential advances have only been observed in the electronics, communications, and bio-chemistry fields, and that the rate of technological progress in most other industries is slowing. What is your opinion?

Technological advance overall is composed of many individual advances, each of which involves a cluster of discoveries and techniques.  By narrowing the field enough you can always find plateaus between advances.  Another apparent damper on advance is satiation -- people are happier having office jobs and cars than they were being farmers and (mostly) walking -- so less effort is expended changing our mode of life.  Basic living arrangements changed less 1950-2000 than they did 1900-1950. So where is the effort expended?  Medical is a major area -- and electronics, and information processing, to name some others; the bulk of what we know in these fields was discovered after 1950.

In the areas where there's interest and effort, progress is accelerating, and there's a cross-fertilization effect as well. So overall I'd agree with Kurzweil, even though some areas appear to be languishing.

Question 9: Many writers and scientists argue that the we will soon see the advent of Artificial Intelligence. Others claim that we are no nearer to truly cognizant, sentient machines than we were 30 years ago. What is your assessment of the prospects for AI?

Oddly enough, 30 years ago is about when I first studied AI.  It was considered a PhD thesis to write a program that could do symbolic integration.  They didn't have any inkling how hard most of the real problems, e.g. vision, speech, robotics, were.  Now they know, and they have a host of techniques they didn't have then, and you can get a computer for $1000 more powerful than any computer in the world was then.

I expect AI somewhere in the neighborhood of 2010.  Alan Turing (who predicted it in about 2000) was a very smart chap, and if the
people who've been working on AI since 1950 had been as smart as he was, we'd have it now.

Question 10: Some writers and scientists argue that the advent of true AI will quickly lead to the development of molecular nanotechnology, and vice versa. Do you agree?

Change those words to "computer modeling and design" and "nanoscale science and technology" and you have a process that's going on right now.  Look ahead just one decade...

Question 11: How much longer do you believe Moore's law will continue? Do you believe that we will soon see molecular electronics?

The physics allows for FET's (field effect transistors, such as current chips use) right on down to the molecular scale.  The only discontinuities are in fabrication technologies; this means there could be plateaus while fab tech catches up, or jumps where a new technique is brought online.  On the average, Moore's law is good for another decade at least, and there's no qualitative separation between "conventional" and "molecular" electronics at the end of that time.

Question 12: What are your plans for the future?

In the short term, I'm writing a book about ethics for robots. In the long term, it probably depends on whether anyone paid any attention to the book.