Fallible humans seeking infallible machines
The lure of the contemporary
‘Imperfect world’ is the canvas on which these notes are painted. This conveys that the world we are part of, and acting in, is one made by people. Insofar as humans are wonderfully and unavoidably flawed, the world we collectively constitute will echo and amplify these limitations. It could not be otherwise.
Another reason for this framing comes from the quixotic desire it holds within it, the promise of perfection. Shaping these notes is a recognition that meaning often resides in the interstices of experience: the gap between the real and the ideal, the is and the ought. In this rendering, for better and worse, both the lure of perfection and our inability to reach it powerfully shape our world.
One response can be found in the Japanese concept of wabi-sabi, which embraces imperfection. A manifestation of this mentality is ensō, a circle drawn in one stroke, which can either be closed (perfection) or open (imperfection). The dynamic is conveyed in the movement embodied in this action, and the practice of repeating it endlessly. Interestingly, there are points of overlap with the eclectic version of Christian thought that Tolstoy developed. He put great emphasis on movement, the possibility of reaching closer to perfection: ‘moral perfection is the impossible goal, but moving to it is the law of human life.’ These different, related ways of thinking about imperfection point towards a humble, fallible mode of engaging with the world.
Another way to approach perfection is to regard it as something realisable. An interesting perspective on the dangers inherent in this temptation can be found in Hannah Arendt’s work on totalitarianism. Reflecting on the machinations of these movements, Arendt observed that, ‘the two great obstacles on the road to such transformation are the unpredictability, the fundamental unreliability, of man, on the one hand, and the curious inconsistency of the human world, on the other.’ There is an attempt to force reality to conform with ideas, to fit the square peg into the round hole. Within this framework, ‘everything which stands in the way of this kind of reasoning – reality, experience, and the daily network of human relationships and interdependence – is overruled.’ Ideas attempt to push down on reality, with brutal consequences. A perfect world is one real people could never inhabit, it requires an impossible degree of consistency and conformity.
To hope for perfection while recognising the impossibility of its attainment is a different endeavour from pursuing it as if it were possible. Demanding perfection is not, however, a problem unique to totalitarianism, although it represents a logical extreme. One of the defining and fundamental errors of that movement was, as Arendt surmised, ‘nothing human is that perfect’. This observation is of enduring relevance, and can be applied more broadly when encountering grand plans and designs. It has again been on my mind as I have been reflecting more on the role technologies are playing in shaping our world and the futures open to us.
On some level, visions of the singularity and the dream of general AI hold within them the desire for perfection. It seems there are dangers either way. If some kind of general AI is actually achieved, then we are left with Arendt’s problem – how to square perfect machines with imperfect humans. Some things – or some ones – have to give. Even if we are able to avoid the apocalyptic scenarios that movies and scholars like to contemplate, there is still much at stake.
One of the fields used to gauge advancements in the capacity of computing has been in the realm of games, most notably Chess and Go, where machines have steadily beaten their human opponents. While it is unclear what these outcomes tell us about machine intelligence, what is more apparent is the damage done to the human experience of mastering these games. Peter Sagal explains how, ‘devotees of the game stake their claim that their pastime is a pure expression of ineluctable human creativity, and then, as the programs improve, the players are stripped of their illusions.’ There is something profoundly sad about this, a situation in which ‘the game they thought was an art is just another mechanism.’ Infinite games are rendered finite, and for what benefit? Here I recall a quote from Walter Benjamin that I have used before:
We have become impoverished. We have given up one portion of the human heritage after another, and have often left it at the pawnbroker’s for a hundredth of its true value, in exchange for the small change of “the contemporary”.
The more likely scenario, however, is that imperfect humans continue to make imperfect machines. Beyond the great dangers that come with Icarus-like projects, the bulk of the problems come with all the minor alterations that cumulatively lead to significant changes. Without necessarily willing it, we might find ourselves in a less forgiving world. Perhaps engineered in certain kind of way, but not necessarily for us.
This predicament of fallible humans seeking infallible machines is potentially one the most consequential aspects of the epoch we are stumbling towards. In finding our bearings, there might be value in better appreciating the human and humane features of imperfection, and the terror latent in hubristic beliefs that perfection is realisable.
This is the second in a series connecting the themes of ‘imperfect notes’ to thinking about the role of technology in society, supported by a grant from the Toshiba International Foundation.