Several researchers were quoted in an ABC News article as being quite skeptical about our timeline for molecular manufacturing. Our claim is that revolutionary capabilities will probably be developed, one way or another, by 2020 if not earlier. They said it would take at least several decades more than that.
So why am I so confident?
One reason is that I believe I've done more thinking about the system-level issues than the timeline skeptics. They have jobs and research projects--they probably have not had time to think in detail about how various capabilities can synergize and make a lot of requirements get easier in tandem.
Another reason is that a lot of the timeline skepticism comes from academia. Academic researchers have several reasons to expect things to take a long time. Perhaps the biggest is that they--at least the good ones--are on the cutting edge, where technologies do not yet synergize. They have to build tools to make tools to make tools, and they have to not just learn, but develop techniques almost every step of the way. Add to that the bureaucracy and grant-writing, and the desire to train new students on the job, and you have a recipe for delay.
Academics in today's funding environment often can't admit when they're working toward something really cool and speculative--they have to take it one tiny predictable step at a time, or they won't get funded. So it is often hard to move quickly toward exciting stuff. And finally, academics may have trouble collaborating (for a number of reasons, some good and some bad), so they may have to reinvent things that another group might have supplied more easily. If molecular manufacturing were going to come out of academia, I would indeed expect it to take 30-60 years.
DNA researcher Ned Seeman raised the point that "Most of the basic principles have not been demonstrated, much less in a 'desktop' context." This is true... depending on what you mean by "basic." For example, low-friction, low-wear, nanoscale linear, rotational, and 2-D sliding bearings have been demonstrated. They were not made by mechanosynthesis, but by manipulating and modifying existing molecules of buckytube and graphite. If mechanosynthesis will be able to make graphite and buckytubes in general-purpose shapes, then it should be able to make bearings. In fact, mechanosynthesis via scanning probe has been demonstrated--but not synthesis of diamond, yet.
So what will it take to pull together buckytube bearings (and what they imply about superlubricity in general), scanning probe chemistry (including reactions that haven't been invented yet), and the other dozen capabilities that will be needed for a nanofactory? I would argue that the biggest single factor is trying explicitly to work on a nanofactory--something which is currently impossible in academia, and unlikely in business as well. Where in academia do you see a team of 100 hand-picked researchers in a dozen different disciplines, all working toward the same goal, all expected to cooperate, and all well-funded with no paperwork? In business, you can get most of those conditions, but it's hard to get the dozen different disciplines in a single company.
Let's not forget that breakthroughs happen. Go back in time just a few years--say, to 2002--and ask any DNA researcher: Would you believe that within five years, there will be a breakthrough that will enable a single novice person, using skills and supplies obtained over the Internet on a hobbyist's budget, to design and make an engineered DNA shape with 10,000 bases in less than a month? I doubt that a single researcher in the world would have thought that was likely. But it has happened.
I can't predict what breakthroughs will happen in the next ten years. Perhaps floppy DNA scaffolds will turn out to be useful for holding buckytubes in a 3D arrangement that can be welded together (and the DNA dissolved simultaneously) with a blast from an electron microscope. Perhaps someone will discover a polymer with many of the properties of protein, but it works just fine in vacuum. Perhaps Zyvex is about to come out with a breakthrough MEMS-based scanning-probe device that will allow high-throughput mechanosynthesis experiments. None of these is likely--but it is likely that there will be one or two dozen equally useful breakthroughs by 2015.
Perhaps my position can best be summarized thus: If, for some reason, a nanofactory project has not started by 2015, but enabling technologies contine to develop and paradigms continue to shift at their current rate, then I would not expect myself to be worried in 2015 that a nanofactory will be unlikely by 2020.