Foresight Update 36
page 4
A publication of the Foresight Institute
 Institute for Molecular Manufacturing Report
The portion of Update 36 that constitutes the IMM Report is on the IMM Web site: http://www.imm.org/.
Book Review: The Robot Age
Foresight involves thinking about the future. Attempting to look out over the twenty-first century, one finds a good vantage point by standing on the shoulders of giants. Two particular giants have recently published books which are in my opinion indispensable to the serious student of the near future: Ray Kurzweil's The Age of Spiritual Machines: When Computers Exceed Human Intelligence (Viking, 1999), and Hans Moravec's Robot: Mere Machine to Transcendent Mind (Oxford, 1999). Both are follow-ups to previous books published about a decade ago: Kurzweil's The Age of Intelligent Machines, and Moravec's Mind Children.
Both authors are intimately involved in the development of the technology they write about. Kurzweil is responsible for a series of ground-breaking systems in text-to-speech, speech-to-text, and related fields; Moravec is a major figure in robotic navigation and vision. The books reflect their specializations to some extent, as they should; read them both for a broad understanding of artificial intelligence, expert systems and robotics. Robot has more about, well, robots, including (recent) history of development efforts, their current status, and likely capabilities in the near future. Age has a broader viewpoint, ranging from computer poetry and investment through translating telephone conversations to virtual reality. There is even a page about me and Utility Fog (p 145).
Moore's Law observes that processing power available per dollar seems to double every year (although the rate of increase is itself increasing). The basic thrust of both books is that relatively early in the next century, the processing power of affordable computers will equal, and then surpass, that of the human brain.
What is the processing power of the human brain? Moravec is perhaps best known for considering the problem in depth in Mind Children. His estimate of 10 trillion operations per second (10 teraops, updated to 100 teraops in Robot) is based on a consideration of the best programs for specific tasks that can be related to specific, reasonably well understood nervous system functions (such as the visual cortex). Thus the estimate implicitly takes account of the serial/parallel, digital/analog, and similar issues. Kurzweil's estimate, 100 times as high, seems to be based on a more direct simulation of brain activity at the neural level. However, this simpler model allows him to propose special-purpose hardware (in the vein of IBM's Deep Blue, but for neurons instead of chess) with the result that both books predict the arrival of $1000 human-level computers sometime in the 2020's.
Here I would take issue with both authors on a point of relevance. Both spend some time discussing the Kasparov-vs.-Deep Blue match. Deep Blue, with about 3% Moravec's estimate of human processing power, is clearly a world-class chess player. There are many human activities which are economically important which do not need the full human physical-world competence. One example is programming.
My workstation's gigaops of processing power just about matches that of a human retina, preprocessing visual images at 10 per second. A programmer's retina, and visual cortex, spends its time converting stuff from computer displays back to some internal representation. (See the "Pigs in Cyberspace" section in Robot.) A software system which did programming wouldn't have to do that (or be able to walk, eat, or tie its shoes). It would still need to be able to understand English well enough that you could tell it what program to write. Even so, it doesn't seem unreasonable that a Deep Blue-class machine could have human-class performance in programming as in chess, given a general-purpose instead of special-purpose computer.
It costs about $100K/year to hire a good programmer (including overhead, taxes, benefits) so a business would be willing to spend about a million dollars (the net present value of that payout) for a programming robot. If we change the model to reflect a 3 teraops machine and million-dollar price tag, we are talking about 5 years from the present instead of 25. All we need now is the program.
Both books talk about programming. Kurzweil goes over three high-level paradigms for AI; recursive search, neural networks, and genetic algorithms. Moravec goes into more detail about algorithms for a robot to form a coherent model of its environment. I would have preferred more analysis of software from each instead of discussions of the origins/fate of the universe (present in both!).
Why think about programming as opposed to the other things semi-intelligent computers can do (and are beginning to do, such as e-commerce)? Although both books note the role of faster computers in designing even faster computers, neither made much of the role of more complex software in developing even more complex software. This is a real trend, not perhaps so noticeable as Moore's Law, but just as important. It is clearly just as much a reflexive capability (see my previous column, "Reflexive Capabilities," in Update 35), a self-referential, self-accelerating loop, as has occurred with computing hardware. With the potential to get to intelligent software but with enormous economic benefits on the way, this could portend the same kinds of major investment that has fed hardware's meteoric rise. If the time scale is as short as I think it may be, the time to start thinking about it seriously is now.
Media Watch.36
by Richard Terra
Technology Review, which bills itself as "MIT's Magazine of Innovation," ran a special report on nanotechnology in its March/April 1999 issue titled "Nanotechnology: The Hope and The Hype." Three separate articles were presented. While it's welcome to see extensive coverage of the field, both the TR editors and the authors of the articles take the unfortunate stance that any work that can't be verified through current laboratory experimental methods is "hype" and "speculation." In an editorial lead to the special report, the editors lump both advanced theoretical modeling of nanomechanical systems and straight-forward extrapolation of potential applications as the imaginings of "nanoaficianados" whose "speculations may sound like the plot of a campy sci-fi movie" that "come primarily from the imagination of K. Eric Drexler," who "has captured the imagination of a loyal band of enthusiasts . . ."
The general tenor of the articles is summed up by the TR editors: "These concepts can be shrugged off as excesses that have done little damage and in fact may have helped fuel public interest in nanotech. Many leading researchers in the field take an indifferent attitude toward these hyperbolic statements, simply pointing out that there is no experimental evidence that such things are possible. But some experts argue that these nanofantasies have taken on a misleading momentum of their own, hijacking the public's fledgling perception of the nanoworld."
The major article, "Will the Real Nanotech Please Stand Up?," by David Rotman (a TR Senior Editor), presents the work and views of a number of nanotech-related researchers, including Richard Smalley, Mark Reed, Phaedon Avouris, and James Gimzewski. These are held up as doing "real" nanotech, because it's rooted in current experimental work. Describing theoretical and advanced design modeling studies by Ralph Merkle at Xerox PARC and Drexler at IMM, on the other hand, Rotman says, "It's somewhere around here that the science starts to become mixed up with science fiction."
Ralph Merkle has posted his letter to TR in response to this article, as well as a more extensive analysis of the article's many flaws, on the Foresight web site at http://www.foresight.org/hotnews/1999MITRev.html, in which he calls for a more balanced view of the value of both experimental and theoretical approaches to the development of molecular nanotechnology:
"We know the goal, we know it's feasible, we don't know how long it will take or exactly which approach is best. Let's investigate the subject: theory and experiment will both contribute, and neither should be denigrated. Each has its particular strengths and its particular weaknesses. By building on the strengths of both, we can develop this new technology more quickly."
A second article, "Moses of the Nanoworld" by freelancer David Voss, profiles Drexler as a "popularizer of nanotech" engaged in "nanotech culture wars" leading a "contingent, which includes computer scientists, technology buffs and believers in cyronics; on the other side is the community of mainstream researchers in physics, chemistry and materials science." The article attempts to show that Drexler, his ideas and his followers are probably irrelevant to the "burgeoning field of nanotech." Voss concludes with the dubious idea that as the nanotech field grows bigger and more inclusive, "it may have left its conceptual progenitor [Drexler] behind." He does not seem to have considered the possibility that Drexler is simply waiting, with some understandable impatience, for the experimentalists in the field to catch up with him.
The third article, "Chips Go Nano," focuses on efforts to push the limits of semiconductor manufacturing technology (UV, X-ray photolithography, etc.) to keep up with Moore's Law.
Technology Review's editorial stance (or at least that of Senior Editor David Rotman) was presaged in an article that appeared in the November/December 1998 issue, titled "Nanotechnology: The Art of the Possible" (available on the web at http://www.techreview.com/articles/nov98/whitesides.htm). Primarily an interview with George M. Whitesides, a chemist at Harvard University, the article opens with this editorial aside: "It's a rapidly developing field with immense potential . . . But it's also an area prone to overblown promises, with speculation about nanomachines that are more likely found in Star Trek than in the laboratory."
Whitesides describes nanotechnology as "an extension of microtechnology" and sees the chief near-term benefits of such research in information storage technology. He talks a bit about microelectronics and UV/X-ray photolithography. When asked about "using nanotechnology for other, mechanical, types of applications," Whitesides responds, "There are a lot of things that range from being potentially real to things that are science fiction. There's the idea of very small autonomous machines . . . I can see no way of realizing those" and lists several objections to nano-scale devices. (Ralph Merkle, ever alert, responded to these with a letter that appeared in the January/February issue).
To be fair, later in the interview, Whitesides responds to a question about the lessons biology holds for nanotechnology, saying, "Biology makes all kinds of very functional small structures. Drexler talks about small motors; we've got a terrific example of a small motor in . . . the flagellar motor in bacteria" and asks, "Can we either learn how to use these biological things in some appropriate way in our devices, or understand the principles of biology to better learn how to embed these principles in non-biological systems?"
So we learn that even level-headed, mainstream laboratory experimentalists occasionally engage in speculation. Technology Review's recent stance equating long-term theoretical studies of nanotechnology and its impacts in general, and the activities of Foresight and IMM specifically, with hype and speculation is puzzling, unfounded — and inconsistent, given their publication of Ralph Merkle's article ("It's a Small, Small World," available at http://www.techreview.com/articles/fm97/merkle.html) in the February/March 1997 issue.
A very brief but much more positive view of Eric Drexler's contributions to the field appeared in a special issue of TIME Magazine titled "The Time 100: Scientitsts and Thinkers of the 20th Century." The citation ("The Engines of Creation," 29 March 1999) says, "When the history of human civilization is rewritten a few centuries hence, the name Eric Drexler just might appear alongside those of Einstein and Freud. Drexler, 43, is the founding father of nanotechnology, the idea of using individual atoms and molecules to build practical machines." The piece concludes: "Today nanotech researchers speak not of if but of when. Great leaps forward come from thinking outside the box. Drexler may be remembered as the man who saw how to build a whole new box."
An article in the February 1999 issue of Business 2.0 ["Let's Get Small," by Glenn McDonald] presents a more balanced view of current research — both theoretical and experimental — and long-term possibilities. The article quotes Ralph Merkle and Foresight Executive Director Chris Peterson, as well as Jim Von Ehr, founder and CEO of Zyvex and a long-time Senior Associate.
Von Ehr's no-nonsense approach to both the promise of nanotechnology and the challenges that need to be overcome to achieve that promise comes through clearly. "While Von Ehr is very careful to distinguish between current short-term initiatives and the more fanciful long-term potential of nanotechnology," the article says, "he's delightfully matter-of-fact about the theoretical possibilities of building things from the atom up."
But Von Ehr's vision of the ultimate goal is equally clear:"'The promise of nanotechnology is as significant as the Industrial Revolution,' says Von Ehr. 'It's bigger than the Internet, bigger than the transistor.'"
Both Merkle and Von Ehr discuss the hard work that remains to be done to realize molecular nanotechnology, and Von Ehr describes the challenges this way: "He outlines three basic problems: First, we must develop chemical processes that enable atomic bonding precisely where and when we want it, and nowhere else; second, develop means for atomically precise positional control . . .; and finally, build a machine that can house both processes, move raw materials in, and move finished materials out."
The article ends by pointing out that the development of advanced molecular manufacturing will bring a fundamental shift toward an economy where information and creativity will loom larger than materials-processing: "In other words, looked at in the finest kind of light, nanotechnology could represent a tidal shift in the basic direction of human business endeavors — away from physical manufacturing into a more notional realm."
(Note: Business 2.0 is online at http://www.business2.com The text of the article is available as a PDF file: http://www.business2.com/articles/1999/02/pdf/vision-a.pdf)
An online article ["Nanotechnology? Make It So!"], based on an extensive interview with Ralph Merkle, appeared 18 February 1999 on the NBC News website, and is still available at http://www.msnbc.com/news/216526.asp. While it presents nothing that would be startlingly new to readers of the Update, it's a good presentation of basic nanotech-related concepts for a general audience. Merkle, as always, is highly articulate in presenting a careful but positive discussion of the promise and potential hazards of molecular nanotechnology. The article is enhanced by links to the transcript of the complete interview, as well as two video clips of Merkle speaking, allowing browsers to experience the full-strength version.
A less successful — and astonishingly badly written and edited — article appeared on the ABC News website dated 31 December 1998 ["Science of the Very Small: Nanotech Visionaries Promise wonders"], and is still available at http://www.abcnews.go.com/sections/tech/CuttingEdge/nanotech981231.html. A bit breathless and poorly focused, the article still manages to present some interesting quotes from Merkle, Von Ehr and Don Noid, a senior scientist engaged in nanotech-related research at the DOE's Oak Ridge National Laboratory. But it's also marred by incomprehensible errors such as the misspellings of Von Ehr's name ("Von Her") and Zyvex ("Zybex"). Hopefully, interested browsers followed the available links to the Foresight, Zyvex, and Merkle's web pages for better, more in-depth presentations.
Research involving carbon nanotubes ("Buckytubes") received extensive coverage in the 11 January 1999 issue of Chemical & Engineering News ["Much Ado About Nanotubes," by Ron Dagani]. The article reports on the meeting of the Materials Research Society held in December 1998. While the article focuses on fullerene nanotube production, purification and bulk chemistry, the article devotes several paragraphs to a presentation by Nobel Prize-winner Richard E. Smalley, Director of the Rice University Center for Nanoscale Science and Technology.
"At a special evening plenary session that attracted an overflow audience, Smalley marveled at the physical properties of single-walled 'buckytubes,' which are being intensively studied around the world. . . . [Smalley] focused his talk, appropriately enough, on the big picture, discussing how buckytubes may help meet the challenges of the next century. At the outset, he urged his listeners, in seemingly contradictory fashion, to 'think nano, think big.' He was referring to the dream of constructing macroscopic objects with nanoscopic precision. Revolutionary new nanotechnology involving carbon-based materials and devices will be crucial to meet the challenge of a burgeoning population, he said. This new technology will be necessary to make the lives of the more than 6 billion people on Earth healthy and fulfilling, he added."
Presentation of the basic goal of developing nano-scale systems, as well as the effort Smalley makes to place that development effort into the context of the enormous socioeconomic challenges we face in coming years, by a well-respected figure before such an audience is a noteworthy and positive event.
The CE&N article also describes an agreement signed last October between NASA and Rice University to "collaborate on the development of new nanotube technology." Initially, the effort will be aimed at producing bulk quantities of fullerene fibers, cables and membranes. But: "The ultimate fruits of this collaboration, Smalley said at the [October] signing . . ., would include 'nano-computers of vastly greater speed and power in dramatically smaller packages than possible with current silicon-based microtechnology,' not to mention new batteries, fuel cells, solar energy devices, composites, sensors, and nanomachines." One wonders what the editors of MIT's Technology Review would make of a Nobel-winning experimental scientist talking before an audience of hard-nosed materials researchers about . . . advanced nanotechnology of the sort discussed by Drexler, Merkle and others.
The Sixth Foresight Conference held last November received coverage in a number of publications: Two papers presented at the conference received an extensive write-up in the "News Focus" section of Science ["Borrowing From Biology to Power the Petite," Science, 283: 27-28 (1Jan99)]. The article, by Robert Service, takes a very positive, nuts-and-bolts approach in describing the work of Carlo Montemagno and his co-workers from Cornell University to study and adapt the protein ATPase to do useful work as a molecular rotary motor. Their paper is available at the conference archive on the Foresight web site at http://www.foresight.org/Conferences/MNT6/Papers/Montemagno/index.html. Also discussed is the work by Violet Vogel and her colleagues at the University of Washington in Seattle with kinesin/tubulin complexes to create simple systems that impart linear motion. An abstract of their presentation is available at http://www.foresight.org/Conferences/MNT6/Abstracts/Vogel/index.html. The potential nano-mechanical applications of both systems are briefly noted.
Research by a collaborative team from Zyvex and Washington University in St. Louis into using an AFM apparatus to probe the properties of buckytubes, as well as manipulate them, was discussed in two separate articles. The team's paper is available at the conference archive on the Foresight web site at http://www.foresight.org/Conferences/MNT6/Papers/Yu/index.html.
The first article appeared in Science ["AFMs Wield Parts for Nanoconstruction," 27 November 1998]. Other scientists involved in nanotube research found the presentation impressive: " 'This is very, very beautiful work', says Anupam Madhukar, a materials scientist at the University of Southern California in Los Angeles. Cees Dekker, a nanotube researcher at Delft University of Technology in the Netherlands, says the most exciting demonstration was one showing two probe tips tugging on a nanotube until it broke. 'That means they can now determine exactly how strong the nanotubes are,' which has long been a goal of the community, he says."
The second article, in Chemical & Engineering News ["Nanotube forced into a jerky form of dance," 23 Nov 1998], covers much of the same ground, but also mentions the conference presentation by Donald W. Brenner at North Carolina State University on the special properties of kinked or crimped nanotubes (the abstract is available at http://www.foresight.org/Conferences/MNT6/Abstracts/Brenner/index.html).
Work by the Washington University team presented at the conference was also briefly mentioned in New Scientist ["Tie a knot in a nanotube," 21 November 1998].
A brief note in Physics Today ["Germany Assembles Nanotechnology Networks," December 1998] describes the startup of six nanotechnology R&D centers in Germany "as part of a DM 150 million ($90 million) push to keep the country internationally competitive in this burgeoning [That word again! -Ed.] field." The research efforts of the centers will focus on ultrathin functional layers, applications of nanostructures in optoelectronics, development of lateral nanostructures, chemical functionalization of nanostructures, ultraprecise measurements and nanostructure analysis. The centers will be organized in a decentralized fashion, drawing talent from academic, government and industry sources to bridge the gap between fundamental research and industrial applications. Each center will receive DM 500,000 per year for three years, but "are expected to seek additional, nongovernment funding and to produce spin-off companies. The hope is that they'll become self-sustaining after five years."
A review/essay of David Brin's nonfiction book, The Transparent Society: Will Technology Force Us to Choose Between Privacy and Freedom?, by Chris Peterson appeared in Reason ["Nowhere to Hide," October 1998]. As she sums up, "Brin's book succeeds in being provocative and stimulating, as advancing technologies threaten to jerk us back and forth between danger and safety, domination and freedom." She invites readers to participate in the online discussion via Foresight's http://crit.org web site.
From Foresight Update 36, originally
published 30 March 1999.
|