Ray Kurzweil’s book “The Singularity is Near” is twenty years old, and its thesis has become conventional wisdom in Silicon Valley. The Singularity is an event horizon – a date at which technological growth becomes so rapid that to look beyond it becomes quite unknowable to pre-Singularity humans, a point at which machine intelligence surpasses human intelligence and goes into a recursive cycle of self-improvement. Kurzweil’s target date for the Singularity was 2045, and in the opinion of many in Silicon Valley we’re well on schedule.
The evidence for the accuracy for Kurzweil’s prediction is, of course, recent rapid progress in AI. But that’s not the only technological development that Kurzweil’s prediction depends on. The connection between machine super-intelligence and control over the physical world needs to be established through nanotechnology.
This needs to be the radical version of nanotechnology as sketched by K. Eric Drexler, in which matter is effectively digitised. In this new world of nanomanufacturing, materials and devices of arbitrary complexity could be assembled atom-by-atom, under software control. As Kurzweil put it, “the revolution in nanotechnology will ultimately enable us to redesign and rebuild, molecule by molecule, our bodies and brains, and the world with which we react”.
Kurzweil’s expectation in 2005 was that “full molecular nanotechnology” would arrive around 2025, a few years before the arrival of superhuman artificial intelligence in 2029. Opinions can differ about whether today’s generative AI is on the road to superhuman AI, but no-one can doubt the huge progress that’s been made in AI in the last twenty years. In contrast, Drexler’s dream of “the principles of mechanical engineering applied to chemistry”, to yield a new form of atomically precise manufacturing, a radical version of nanotechnology, remains largely unrealised. What happened?
Explanations of this slow progress fall into three categories – the political, the practical, and the conceptual.
In the view of many proponents of the original vision of molecular nanotechnology, it was blocked by a conservative scientific establishment that feared disruption, and preferred to divert resources towards more conventional materials science. While it’s true that there were some bad tempered and unnecessarily ad-hominem debates in the early 2000s, I don’t think this argument is convincing.
Firstly, it misunderstands the decentralised nature of science: distinguished elder scientists are influential, but proving Nobel Laureates wrong is a great route to career success. Secondly, it is a rather parochial view – these were debates in the US science community that weren’t binding on the rest of the world, and the wider field of nanotechnology was one that the USA didn’t dominate then, and does so even less now. If developing molecular nanotechnology was easy, why would we think that it wouldn’t have been done by now in China?
In the second point of view, the early proponents of molecular nanotechnology simply underestimated how hard in practise it would be to make their ideas work. In fact, outside the USA, the debate around molecular nanotechnology was much less heated. For many experimental scientists in related fields, the ideas were exciting, but the largely theoretical work of Drexler and his followers had simply underestimated the practical difficulties. Back in 2005, I identified six practical challenges that, in my opinion, stood in the way of developing molecular nanotechnology, and the research that would be needed to overcome them. I don’t think a lot of progress has been made in addressing these and other issues since then.
A third line of argument returns to the lessons to be drawn from cell biology. As Drexler convincingly argued, the molecular machines and intricate structures that we see in cell biology provide an existence proof for a sophisticated nanotechnology. But what does biology tell us about the best way to create a synthetic analogue?
We can get some insight into why the original Drexlerian vision didn’t progress, by looking at a couple of other approaches to creating structures and devices with atomic precision, which have at least progressed as far as laboratory demonstrations, if not mass application. The supramolecular chemistry approach to molecular motors won a chemistry Nobel prize for Ben Feringa in 2016, while the persistence and vision of the late Ned Seeman founded the field of DNA nanotechnology, which has, in a restricted domain of material types, achieved a version of digitally specified molecular-scale structures and devices. These approaches, in different ways, learn from how biology works at the nanoscale – but their operating principles look very different from the mechanical engineering inspiration of Drexler.
The key point here is that the physics that is operative in nanoscale biology, in the warm wet world of the cell, looks very different from the physics that rules at the macroscale. It’s dominated by Brownian motion, surface forces are very strong, and the watery environment is dominated by viscosity, with inertia being essentially negligible. Cell biology uses entirely different design principles that are optimised for this world, with mechanisms such as self-assembly and molecular conformational change that have no counterparts in the macroscopic world.
This is especially relevant for one application of radical nanotechnology that is central to Kurzweil’s vision – the idea of tiny nanobots navigating the bloodstream fixing the body cell by cell. It’s these nanobots that Kurzweil argues will be able to read out the state of a human brain, permitting the “up-loading” of the mind to super-powerful computers. This provides the route to personal immortality that seems so important to Kurzweil and his followers.
The first realisations of some approximation to the nanobot vision were in cancer treatment, where the idea of selectively delivering chemotherapy agents to cancer cells led to much research, and a few applications, based on biologically inspired principles of self-assembly and environmental responsiveness. It was the covid epidemic that led to the coming of age of this line of research; it provided the delivery mechanisms for messenger RNA in the new vaccines from Moderna and BioNTech. But these systems are very much “soft machines”, working by analogy with cell biology, rather than using Drexlerian nanotechnology.
Given the lack of progress in molecular nanotechnology, does that mean that the Singularity must be postponed? That’s clearly not the Silicon Valley view; for the evangelists, the imminent arrival of artificial general intelligence will solve everything, including rapidly developing molecular nanotechnology. If it really was just politics that caused the slow progress to digital matter, then that might be a plausible argument. But if the problem is more fundamental, arising from taking the wrong lessons from biology, then it’s not clear that AI will help.