The Unique Character of MNT Exploratory Engineering

From Wise Nano

Any serious attempt to predict the future must ultimately confront the abject failure of typical technological projections. Skepticism does not seem to provide much protection, as for every technophile's promise of imminent nuclear power too cheap to meter or conversational AI in the 1960s there seems to be a president of the royal society Lord Kelvin declaring "heavier than air flying machines are impossible" in 1895, 8 years before their invention or an Albert Einstein claiming that "There is not the slightest indication that nuclear energy will ever be obtainable. It would mean that the atom would have to be shattered at will" in 1932, 13 years before Trinity.

In the face of such precedents, how can we hope to plan for the future at all? One question that should surely be asked is what current decisions would be "wise". If we were to plan for the near-future development of MNT, and it were to turn out that MNT is surprizingly difficult and is not developed in the next 50 years, what would we loose? This depends on what precautions we took. Failing to explore alternative energy sources because of confidence in MNT solving current energy crises would have adverse consequences. On the other hand, most of the proposed measures for preparing for MNT would be desirable in any event. After all, few people seriously deny that radically increased technological and particularly productive capabilities will emerge over the course of the century. The development of economic preparation for MNT driven production will also be of value in response to rapid productivity increases from any other cause, and of some use in response to economic disruption in general. Learning how to prevent biotech arms-races and how to do the same with nanotech are similar tasks. In general, efforts aimed at preparing for the development of MNT are unlikely to be harmful even if the anticipated technology fails to emerge. Likewise, the possibility of failure is no reason to deny funding. All sorts of engineering tasks are attempted, in the well verified expectation that while it is impossible to predict winners ahead of time, those which are funded will more than justify the failures. In fact, serious impediments to technological progress may arise from failures to admit the possibility of failure. For instance, government concentration of hundreds of billions of dollars of research funding on a few programs for the development of nuclear power, rather than spreading this money among thousands of projects exploring a variety of development paths for a variety of alternative energy possibilities may be a major reason for continuing dependence on fossil fuels.

At any rate, there appear to be some strong reasons for believing that MNT will develop according to the expected time-line. The most prominant reason for this is that almost uniquely among projected technologies, the approximate date of MNT development projected by the most knowledgable experts has remained constant, or has actually advanced slightly. By contrast, fusion power, AI, and affordable space travel have maintained a roughly constant distance from development. Even more shockingly, we appear to be about as far from a paperless office or from routine use of voice based computer interfaces as we were almost twenty years ago, despite progress in performance metrics continuing at almost exactly the projected rates. The reasons for this difference are quite simple. Fusion and AI are fundamentally scientific research programs. The first field is hampered by a lack of an adequate mathematical description of the behavior of high temperature plasma. Plowing forward with engineering in the absence of sound science can be successful, but it renders timelines pure speculation. Also, the combination of inadequate science and over concentrated funding for a single approach to a problem practically guarantees failure. The vast majority of AI efforts have long ago ceased to even pursue the goal of producing or even advancing towards real generalized cogniton. What efforts still aim at this elusive goal struggle alone, devoid of even the rough outlines of a consensus theory. By contrast, the well timed benchmark predictions but practical failures of computer technology are due to the failure of futurists to account for human unwillingness to sacrifice any of the valuable features which they are accustomed to in order to transition to a new technology. Also significant is the failure to appreciate the reluctance of non-technophiles to adopt a new technology when the old one is meeting their needs, especially if doing so would add occasional inconveniences to what is currently a routine task. However, it should be noted that in the field of routine advances in miniturization, timelines are predictable. Of course, MNT is closer to a non-routine advance in miniturization, but even these have developed at a predictable rate, keeping Moore's Law stead despite the repeated instances of supposedly fundamental barriers.

Neither of these classes of problems are relevant to the development of MNT. The basic science of manipulating matter on the molecular scale, either enzymatically, through self assembly, or in vacuum, is well understood and has been for decades. MNT researchers display differences of opinion regarding what approach to development is likely to yield the fastest results, but there are no fundamental differences regarding the feasibility of standard proposals. Drexler's work in Nanosystems, and subsequent work by Frietas and Merkle have demonstrated that none of the common critiques of MNT are valid, and that for the most part nanoscale systems can be produced which function without requiring the novel science of deeply understanding nano-scale materials, complex quantum phenomena, etc. These are surely not the best possible nanosystems, but they are sufficiently superior to current products that there is little question regarding the high demand they would generate. The simplification shown by nanosystems, and later by the convergent nanofactory, suggest another fundamental reason for MNT development staying on track. 1970s molecular biology and 1980s probe microscopy provided the basic science necessary for MNT, and early proposals, such as in Engines of Creation, provided a description of a technology that was clearly possible. The technology proposed at the time was definitely capable of working as described. No further discoveries were required to implement an exploratory engineering proposal such as this one, only clever insights into how to resolve the admittedly messy details were needed. Because the initial design was aimed primarily at proving plausibility, not at proposing a simplified system for immediate development, the surprises have consistantly been favorable to prediction timelines. While in AI tasks repeatedly turn out to be more complex than initially expected, and then yet again harder than expected after compensation is made for Murphy's Law, in MNT we know that the initial proposal is complex enough, and the movement has continually been towards designs that exceed the original in simplicity. A good example of this is the fine motion controller, which was initially shown to require no more than a few million atoms, and is now known to require at the most a few thousand. Further progress has been made possible by abandoning, at least for now, goals which were recognized as unnecesary to MNT's economic impact, if not actually undesirable. The most notable examples of this are autonomous replication, and the flexibility to work with all kinds of atoms or molecules. The analogy that leaps to mind is the airplane, which was demonstrated to be possible by the complex aparratus of avian flight, but which was made practical by the expediant sacrifice of winged motion and the attendant unnecessary VTOL capabilities. By thus reducing complexity, progress has been kept on track despite serious disappointments regarding the misappropriation of federal and corporate funding.