Tag Archives: artificial intelligence

rocket from the crypto / two dead letters

Chairman Bruce appears to be repubbing longreads from the now defunct Beyond The Beyond blog. This is a weird experience for me—distinctly atemporal, to use the man’s own term—because I recall reading this stuff at the time. And so it’s familiar and just-like-yesterday, but also so alienated and impossibly historical… I mean, I can’t recall the last time I saw anyone so much as mention the New Aesthetic, but I certainly remember a time when it seemed like everyone was talking about it. (That feeling of atemporal synchronicity is being compounded, no doubt, by my having been going through some of my own published material from the same period over the last couple of weeks… with the added irony that said act of retrospection was to the end of writing a chapter about Sterling for an academic collection.)

TL;DR—middle-age is a headfuck. I kind of understand why my parents went so weird in their forties, now… though I’m not sure I yet forgive the particular direction in which they went weird. And they didn’t even have the internet!

Anyway, the essay in question is the Chairman’s response to the New Aesthetic panel at the 2012 SXSW, and the bit I’m clipping is less about the New Aesthetic than a side-swipe at AI that reads just as true (and just as likely to be ignored) today:

… this is the older generation’s crippling hangup with their alleged “thinking machines.” When computers first shoved their way into analog reality, they came surrounded by a host of poetic metaphors. Cybernetic devices were clearly much more than mere motors and engines, so they were anthropomorphized and described as having “thought,” “memory,” and nowadays “sight” and “hearing.” Those metaphors are deceptive. These are the mental chains of the old aesthetic, these are the iron bars of oppression we cannot see.

Modern creatives who want to work in good faith will have to fully disengage from the older generation’s mythos of phantoms, and masterfully grasp the genuine nature of their own creative tools and platforms. Otherwise, they will lack comprehension and command of what they are doing and creating, and they will remain reduced to the freak-show position of most twentieth century tech art. That’s what is at stake.

Computers don’t and can’t make sound aesthetic judgements. Robots lack cognition. They lack perception. They lack intelligence. They lack taste. They lack ethics. They just don’t have any. Tossing in more software and interactivity, so that they’re even jumpier and more apparently lively, that doesn’t help.

It’s not their fault. They are not moral actors and they are incapable of faults. It’s our fault for pretending otherwise, for fooling ourselves, for projecting our own qualities onto phenomena that we built, that are very interesting to us, but not at all like us. We can’t give them those qualities of ours, no matter how hard we try.

Pretending otherwise is like making Super Mario the best man at your wedding. No matter how much time you spend with dear old Super Mario, he is going to disappoint in that role you chose for him. You need to let Super Mario be super in the ways that Mario is actually more-or-less super. Those are plentiful. And getting more so. These are the parts that require attention, while the AI mythos must be let go.

AI is the original suitcase word; indeed, it’s a term that Minsky came up with to describe the way the goal of “AI” kept drifting, and coming up with the term and identifying the problem didn’t get him anywhere nearer to solving it. I was writing a report on AI last year in a freelance capacity (for a foundation in a location whose commitment to the Californian Ideology is in some ways even greater than that of California itself, despite—or perhaps because of—its considerable geographical, historical and sociopolitical distance from California), and tried to make this point, drawing on the tsunami of critiques of AI-as-concept and AI-as-business-practice that have emerged since then, both within the academy and without… but, well, yeah.

I guess we just have to conclude that the sort of person who decides to make Super Mario their best man is not the sort of person who’s going to take it well when you point out that Super Mario is a sprite… no one wants to be the first to concede the emperor is naked, particularly not when they’ve stripped off in order to join the parade. Nonetheless, given the residual enthusiasm for peddling that particular brand of Kool-Aid which still persists among the big global consultancies, the McKinseys and their ilk, there’s probably a few more years in business models offering “Super Mario solutions” before smarter, faster-moving players start focussing on practical applications without the pseudo-religious wrapper. Or, I dunno, maybe not? Seems like people will believe whatever the hell makes them feel like a winner these days, and the very unfalsifiable nebulousness of “AI” might make it all but bulletproof for that very reason. Every era has its snake-oils.

far from the ideological panoply promised

  1. Aware that the regulations concerning feedback to the Central Civic Authority limit my initial complaint to one hundred sentences, I am indulging in sentence structures more complex and digressive than would be my inclination otherwise.
  2. It is my understanding that the daemon constructed to assess and rout complaints to the relevant authorities is equipped to understand any grammatically correct sentence.
  3. If this is not the case, I request notification of this fact immediately.
  4. As a gesture of good faith toward the one-hundred-sentence regulation, and a demonstration of my commitment to the civic goodwill, I will eschew the use of semicolons throughout this document.

An inventive and subtly pointed piece of fiction from Alex Irvine over at Lightspeed Magazine. Recommended.

a metrics of labour other than time

Very interesting long paper by Matteo Pasquinelli; going back through Marx’s notion of the general intellect, he shows that none other than yer man Babbage theorised computing systems not only as a concretisation of labour but a crystallisation of preexisting biases in the workforce. Everything old becomes new again.

… the distinction between manual and mental labour disappears in Marxism because, from the abstract point of view of capital, all waged labour, without distinction, produces surplus value; all labour is abstract labour. However, the abstract eye of capital that regulates the labour theory of value employs a specific instrument to measure labour: the clock. In this way, what looks like a universal law has to deal with the metrics of a very mundane technology: clocks are not universal. Machines can impose a metrics of labour other than time, as has recently happened with social data analytics. As much as new instruments define new domains of science, likewise they define new domains of labour after being invented by labour itself. Any new machine is a new configuration of space, time and social relations, and it projects new metrics of such diagrams. In the Victorian age, a metrology of mental labour existed only in an embryonic state. A rudimentary econometrics of knowledge begins to emerge only in the twentieth century with the first theory of information. The thesis of this text is that Marx’s labour theory of value did not resolve the metrics for the domains of knowledge and intelligence, which had to be explored in the articulation of the machine design and in the Babbage principle.

Following Braverman and Schaffer, one could add that Babbage provided not just a labour theory of the machine but a labour theory of machine intelligence. Babbage’s calculating engines (‘intelligent machines’ of the age) were an implementation of the analytical eye of the factory’s master. Cousins of Bentham’s panopticon, they were instruments, simultaneously, of surveillance and measurement of labour. It is this idea that we should consider and apply to the age of artificial intelligence and its political critique, although reversing its polarisation, in order to declare computing infrastructures a concretion of labour in common.

Staring down Roko’s basilisk

Pete Wolfendale:

We have consistently overestimated what computation is capable of throughout history, whether computation was seen as an algorithmic method executed by humans, or a process of automated deduction realised by a machine. The fictional record is crystal clear on this point.

Instead of imagining machines that can do a task better than we can, we imagine machines that can do it in the best possible way. When we ask why, the answer is invariably some variant upon: it is a machine and therefore must be infallible.

This is absurd enough in certain specific cases: what could a ‘best possible poem’ even be? There is no well-ordering of all possible poems, only ever a complex partial order whose rankings unravel as the many purposes of poetry diverge from one another.

However, the deep, and seemingly coherent computational illusion is that there is not just a best solution to every problem, but that there is a best way of finding such bests in every circumstance. This implicitly equates true AGI with the Godhead.

Cold equations in the care vacuum

In a nutshell, over-reliance on computer ‘carers’, none of which can really care, would be a betrayal of the user’s human dignity – a fourth-level need in Maslow’s hierarchy. In the early days of AI, the computer scientist Joseph Weizenbaum made himself very unpopular with his MIT colleagues by saying as much. ‘To substitute a computer system for a human function that involves interpersonal respect, understanding, and love,’ he insisted in 1976, is ‘simply obscene.’

Margaret Boden at Aeon, arguing that the inability of machines to care precludes the “robot takeover” scenario that’s so popular a hook for thinkpieces at the moment.

I tend to agree with much of what she says in this piece, but for me at least the worry isn’t artificial intelligence taking over, but the designers of artificial intelligence taking over — because in the absence of native care in algorithmic systems, we get the unexamined biases, priorities and ideological assumptions of their designers programmed in as a substitute for such. If algorithmic systems were simply discreet units, this might not be such a threat… but the penetration of the algorithm into the infrastructural layers of the sociotechnical fabric is already well advanced, and path dependency means that getting it back out again will be a struggle. The clusterfuck that is the Universal Credit benefits system in the UK is a great example of this sort of Cold Equations thinking in action: there’s not even that much actual automation embedded in it yet, but the principle and ideals of automation underpin it almost completely, with the result that — while it may perhaps have been genuinely well-intended by its architects, in their ignorance of the actual circumstances and experience of those they believed they were aiming to help — it’s horrifically dehumanising, as positivist systems almost always turn out to be when deployed “at scale”.

Question is, do we care enough about caring to reverse our direction of travel? Or is it perhaps the case that, the further up Maslow’s pyramid we find ourselves, the harder we find it to empathise with those on the lower tiers? There’s no reason that dignity should be a zero-sum game, but the systems of capitalism have done a pretty thorough job of making it look like one.