The Materiality of Life and What it Means for What the Real Revolution in Artificial Intelligence Will Look Like

Back in the 1980s Alvin Toffler promulgated visions of an "information age" which would see the birth of an information civilization in which work, wealth, consumption and everything else would be increasingly "dematerialized." Material goods would still exist, of course, but the point was that we would become much, much more efficient at extracting, processing and consuming them, as we substituted information for "crude matter." Indeed, reading Toffler's books The Third Wave and Powershift it seems clear that we were supposed to be on the road toward a world where our manufacturing would become so successful in replacing the scarce and expensive and awkward with the abundant and cheap and convenient; and rigidity with flexibility; as to approach an ideal of cost-effectively making anything out of anything else, an object underlined by the bold declaration opening the preamble of the "Magna Carta for the Knowledge Age" he coauthored with like-minded colleagues George Gilder, George Keyworth and Esther Dyson: "The central event of the 20th century is the overthrow of matter . . . physical resources . . . losing value and significance" in economic and political life, for "[t]he powers of mind are everywhere ascendant over the brute force of things."

Alas, from the standpoint of the 21st century's third decade things look quite different from Toffler's vision of a dematerialized information age. As we have been reminded again and again since that supposed overthrow of matter physical resources still have immense, even decisive, significance in economic and political life. We were so reminded amid the commodity price boom that, not least by enabling the recovery of Russian power, helped redraw the geopolitical map, while plunging much of the world into a crisis of food and fuel prices. We were so reminded again as the pandemic, disrupting work processes, underlined the continued importance of persons doing physical labor to produce physical things and render physical services at particular physical locations, with the subsequent price shocks showing how little capacity we still had to adapt to a situation in which those persons did not perform those tasks in the same old ways. We had another reminder in the way that the entirety of the North Atlantic Treaty Organization, with a manufacturing "value added" fifty times that of Russia, failed to match Russian production of plain old artillery shells--showing how far from that perfect fungibility of productive capacity real-world manufacturing remains. And we had perhaps the most fundamental reminder in the way that "overshoot day" came earlier and earlier each year, as the failure of business to become more efficient at using and making goods and services accelerated and deepened the ecological crisis on just about every front.

All this is something to keep in mind as we consider the revolution in Artificial Intelligence (AI) supposed to be imminent, if not already ongoing, the more in as the more astute analysts of information technology have for decades stressed particular areas where the technology has proven of limited utility, and which have thus translated to its effect on the economy being a far cry from the hype. Chief among these is the capacity of machine intelligences to "sense" the world around them so as to navigate it safely and handle objects in it with dexterity, especially in situations where the environment and the task is complex, variable and requires a steady stream of responses tailored to the individual situation--robots useful on assembly lines producing high value added items like cars, but a robot which can do your laundry elusive.

Back at the height of the "tech boom" euphoria Robert Gordon lucidly pointed this out as thus far limiting the impact of the innovations of the "New Economy," and likely to go on doing so. It is a testament to his grasp of the issue that surveying the possibilities of automation many years later after that particular boom of techn-hype went bust Carl Benedikt Frey and Michael Osborne acknowledged these same problem areas as the critical bottlenecks to automation--if in the expectation that progress was being made here, enough so as to trigger something like panic in some quarters. However, a decade on it still seems that this has been an area where, again, progress of the kind they thought was happening has proven slow indeed, exemplified by the failure to produce the really viable self-driving car in the years since that Frey and Osborne seem to have been rather sanguine about in their study, never mind a robot that can do your laundry.

Indeed, it seems telling that the recent surge of hype about artificial intelligence has had nothing to do with any breakthrough producing machine intelligences capable of observing, navigating, manipulating the world around them with human-like versatility, reliability and efficiency. Rather that hype has revolved around Large Language Models--and while at one end of the discourse some people dismiss them as "glorified autocompletes," and at the other some scream that they are Lovecraftian demons to which Silicon Valley has opened the portal, no one denies their limitations in interacting with the physical world. One result is that these artificial intelligences so ill-suited to physical tasks yet (their proponents say) becoming as good as, if not better than, humans at many mental tasks have inverted the cliché about what those "technologically displaced" in the job market must do to go on getting a paycheck--the truck driver not having to "learn to code," but the now superfluous coder needing to learn to drive a truck.

Of course, AI that really does prove to be good enough to replace coders would be consequential, the more in as a stronger artificial intelligence of that type can be used in cases where innovators figure out ways to design the physical activity out of tasks--just as we did with filing. Rather than an office robot that picks through a filing cabinet to get a desired manila folder, computers store electronic files in electronic memories, and it is far from clear that we have done all we can here. (Consider, for instance, how "generative" artificial intelligence may be able to replace physical TV and film production by conjuring up video from a prompt the way Sora is intended to do--a capability that may by the time you are reading this item already have reached the consumer in a crude form.)

Nevertheless, in this material world what will really, really matter is what will happen when--if?--the progress of artificial intelligence reaches the point at which a robot can be counted on to pick up your socks off the floor, add them to the pile of clothes already in the laundry hamper, then take them off for laundering and bring them back to you clean and dry.