A World Without Work
A World Without Work: Technology, Automation, and How We Should Respond by Daniel Susskind (2020)
![]() Passtimes of the future, as imagined by Daniel Susskind
|
Dr. Susskind, scion of the storied futurology dynasty, will doubtless find enough general counsel who are anxious to seem at the technological vanguard — and interested mugs like me, who are suckers for sci-fi alternative histories — at least to recoup his advance, but A World Without Work will not signpost much less dent the immutable trajectory of modern employment, misunderstanding as it does how humans, organisations or economies work, while ignoring — neigh, contradicting — the whole history of technology, from the plough. An excellent counterpoint, though equally flawed in other ways, is the late David Graeber’s highly provocative Bullshit Jobs: A Theory, which has a far more realistic, if no less glum, prognosis: soul-destroying jobs aren’t going away: they are only going to be more and more of them. This feels more plausible to me.
Technology has never destroyed overall labour, and Susskind gives no good grounds for believing it will suddenly start now.
No innovation since the wheel has failed to create unexpected diversity or opportunity — that’s more or less the definition of “innovation” — or more subsidiary complexity & inefficiency as a by-product. Both the opportunities and the inefficiencies “need” human midwifery, to exploit them (for the former) and effectively manage them (for the latter).
Nothing that the information revolution has yet thrown up suggests any of that has changed. The more technology is deployed, the more the fog of confusion and complexity — as in complexity theory, and not just complicatedness — engulfs us.
But chess-playing supercomputers -
Hand-waving about Chess and Go-playing supercomputers — there is a lot of that in A World Without Work — does not change anything. In the world of systems analysis, Chess and Go are complicated, not complex, problems. Both are hermetically and — ahh — hermeneutically sealed zero-sum games on small, finite boards with simple sets of unvarying rules between two players sharing a common and static objective. Their risk payoff is normal, not exponential. They can, in theory, be “brute force” managed by skilled operation of an algorithm, and the consequences of failure are predictable and contained — you lose.
Gameplay is entirely deterministic: you can see that, at the limit, the player with the better number-crunching power must win. Even here, the natural imagination of human players, otherwise at a colossal disadvantage from an information processing perspective, makes beating them surprisingly hard.
This ought to be the lesson: even for thoroughly simplistic binary games, it takes a ton of dumb processing power to beat a puny imagineer.
But somehow, Susskind reads this as a signpost to the Apocalypse.
Look: life is not a two-person board-game on a small-board with fixed rules and a static, common, zero-sum objective. Life is complex. Complex problems — those one finds at the frontier, when one has boldly gone where no-one has gone before, in dynamic systems, where information is not perfect, where risk outcomes are convex — so-called “wicked environments” — are not like problems in Chess.[1] Here algorithms are no good. One needs experience, wisdom and judgment. Algorithms get in the way.
Computers can’t solve novel problems
By design, computers can only follow rules. A machine that could not process instructions with absolute fidelity would be a bad computer. Good computers cannot think, they cannot imagine, they cannot handle ambiguity — if they have a “mental life”, it exists in a flat space with no future or past. Computer language, by design, has no tense. It is not a symbolic structure, in that its vocabulary does not represent anything.[2] Machines are linguistically, structurally incapable of interpreting, let alone coining metaphors, and they cannot reason by analogy or manage any of the innate ambiguities that comprise human decision-making.
Until they can do these things — and conceptually there is no reason a machine couldn’t, but that’s just not how modern computers have been designed — they can only aid, and in most circumstances, complicate, the already over-complicated networks we all inhabit.
But chess-playing supercomputers -
But, but, but — how can we explain this seemingly relentless encroachment of the dumb algorithm on the inviolable province of consciousness? What will be left for us to do? Well, there’s an alternative explanation, and it’s a bit more prosaic: it is not so much that AI is breaching the mystical ramparts of consciousness, but that much of what we thought required the ineffable, doesn’t. Much of what we thought was “human magic” turned out to be just, in Arthur C. Clarke’s worlds, “sufficiently advanced technology”.
This isn’t news: impish polymath Julian Jaynes laid it all out in some style in 1976. If you haven’t read The Origin of Consciousness in the Breakdown of the Bicameral Mind, do. It’s a fabulous book. In any case, a lot less of what we take to require conscious thought actually does require conscious thought. Like driving a car. Or playing the piano.
And even this is before considering the purblind, irrational sociology that propels all organisations, because it propels all individuals in those organisations. Like the academy in which Daniel Susskind’s millenarianism thrives, computers work best in a theoretical, Platonic universe, governed by unchanging and unambiguous physical rules, and populated by rational agents. In that world, Susskind might have a point — though even there, I doubt it.
But in the conflicted, dirty, unpredictable, complex universe we find ourselves in, out here in TV land, there will continue to be plenty of work, as there always has been, administrating, governing, auditing, advising, rent-seeking — not to mention speculating and bullshitting about the former — as long as the computer-enhanced, tightly-coupled complexity of our networks doesn’t wipe us out first.
Employment and Taylorism
Susskind’s conception of “work” as a succession of definable, atomisable and impliedly dull tasks — a framework, of course, which suits it perfectly to adaptation by machine — is a kind of Taylorism. It is a common view in management layers of the corporate world, of course — we might almost call it a dogma — but that hardly makes a case for it.
The better response is to recognise that definable, atomisable and dull tasks do not define what is employment, but what it should not be. The JC’s third law of worker entropy is exactly that: tedium is as sure a sign of waste in an organisation. If your workers are bored, you have a problem. If they’re boring each other,[3] then it’s an exponential problem.
Daniel Susskind does not say how using artificial intelligence to bore each other is going to change that.
See also
References
- ↑ There is more on this topic at complex systems.
- ↑ See: Code and language.
- ↑ Hello, financial services!