Failed prophecies

“Sanctify them in the truth” (John 17:17)

Tesla (the car business) recently published a video of its AI day giving a tour de force of modern artificial intelligence (AI) methods. From monte carlo tree search to simulation-based reasoning, Tesla is throwing every modern technique to the problem of self driving cars, coupled with possibly some of the best research scientists and engineers in the business. Yet, somehow, this is not enough; as another video mocking Tesla for its inability to produce a self-driving car points out, the promise of real-world AI delivering anything that would resemble its science fiction counterpart is astoundingly far. The likes of self-driving cars, robotic maids and butlers, robotic farm work, automated medical doctors, robotic you-name-it are currently a pipe dream, and might remain so for the ages (more on this later). Tesla operates within the mantle of extreme technological optimism, an almost default positions given the level of change the world has seen since the industrial revolution. This level of optimism is not unique to Tesla – the techno-social landscape has been a source of unparalleled optimism across the political spectrum. Living in a world without progress is beyond our ability even contemplate, but this might be exactly what we might be facing.

Artificial intelligence, optimism and failure

It wouldn’t be that unfair to claim that the history of AI is the history of attacking games and mathematical puzzles. The core belief was that if we manage to successfully attack Chess, Go and Poker, together with some innovations in the sensory (i.e. speech, vision) and language pipelines, we would have mastered the necessary technology to create intelligent machines. I consider the original game research programme concluded in spirit (i.e. we now have super-human players games that range from poker to starcraft), but its impact on “real world” is much less pronounced. The core of modern AI is a fusion between logic and statistical learning (often under the moniker of “machine learning”). While machine learning did originate within the computer science community, the scientific foundations were of a statistical bend. Behind a mathematical façade statistics alludes to a core belief, which can summed up as: “we can predict the feature because we can measure the past”. This turns out to be extremely hard outside the confines of mental abstractions. Most of us have heard the story (from school) that you can only add apples to apples and oranges to oranges (i.e., you can only count likes). In the first instance, this seems trivial, but at as you delve deeper distinctions in the world only make sense in terms of their utility. There are myriads of measurements that make up an object (or an event) and setting up categories (implicitly or explicitly) is not easy. Can you add large apples to small apples (hint: yes, if you plan on following this up with a measurement of tonnage)? Green apples to red apples (hint: no if you are selling green apples) and so on and so forth. This demarcation of when a quantity becomes a quality is not trivial at all and opinions on how to approach this successfully vary. In turn, our inability to count properly leads to failure when it comes to “out-of distribution generalisation” – machines fail to extrapolate from data. This is in stark opposition with what you will find in leftist texts on the topic. For example, quoting directly from FALC “…after that, incorporate machine learning which is able to respond to unexpected situations arising beyond the data viewed as otherwise typical”. Machine learning cannot handle unexpected events, it just deals with patterns available in the data.

Humans are so good at generalising in unknown environments that we make universes in our heads that never existed and reason about them (i.e. any fantasy novel); machines fail miserably to account for even minor novelty. Once you add all other problems to the mix (limited sensory apparatus, limited processing capabilities, spurious correlations, lack of meta-knowledge, catastrophic forgetting – the list goes on), we are really far from having a proper solution to the problem of automation. We might not even be approaching the whole setting in the right angle, and any solution might require years of research, turns and twists and bumpy roads before it is made viable.

Interestingly enough, it looks like we have the mechanical hardware, but we are literally clueless on how to replicate human-level intelligence. While this is a somewhat dangerous prediction to put down on paper, to a certain extend AI is a proto-science, like alchemy, and it might take another 300 years to just get a realignment of goals (i.e. no meaningful transmutation, ever!). Beyond AI, most other forms of sci-fi technology are currently not viable either (from your cool cybernetic eyes to “trivial” advances like curing male pattern baldness and cancer). Maybe one can blame the current funding landscape (too much focus on financial exploitation, not much exploration), but I doubt there is a grant (!) conspiracy behind this (on the contrary, billionaire money is flowing towards alchemical goals like immortality), it’s just that the problems are hard and nobody knows where the next big thing will come from. The low hanging fruits have been picked 50-70 years ago and we are not sure where to look next. Billion dollar R&D investments do not seem to be solving the problem.

AI and value

If robotic slaves are not possible (or plausible), what kind of value is AI adding to businesses? Why are big consultancies so eager to get a data science department within each enterprise? Consultancies and tool makers have tremendous incentives to promote a technology that is not mature as they can keep charging exorbitant hourly rates, but surely you would think that the buyers would have seen through the sale pitches? Most businesses are looking with help with sales (the most common data operation has to be “have a look into my data, get me good customers”) and process automation. If you can strip a problem down to its essentials, you can throw away all forms of algorithmic reasoning (which I would be pressed to call AI, though some of it is of a statistical nature). At this point you are doing some form of glorified accounting/administration, and contrary to what your manager will tell you, being the boss is easy. You tell people what to do in vague abstract terms (e.g. pick up person X from street y), they apply their ingenuity to do it (i.e. fight traffic) and you get to penalise/appraise them for it. The princes of this world are not cast out by automation, they are strengthened and multiplied in algorithmic form. Value is still retained within labour (in fact, it could not be otherwise) – the vague algorithmic reasoning is quite often a modernisation show. To make matters worse, tt is not clear if these methods, when applied to something other than very streamlined workflows, are effective in an way. Having an algorithmic manager/consultant is one thing, having one that’s broken creates a new Taylorist hell. Middle-class professions are not immune to broken deployments either, though they have proven more resistant until now. You do not need to go far to see how bad these “intelligent” services are – just try using a random chatbot online and marvel at machine incompetence.

Christianity, liberalism and automation

Christianity (of the churchgoing variety) is losing followers fast, as secularisation takes hold in the major Christian nations. One could argue that secularisation mostly entails liberalism, which, broadly speaking, has two tenets – though its adherents would not necessarily agree with anything apart from being rationalists. The first one is the use of the market as a means of social mediation. The project of liberalism traces its roots back to the European religious wars of the 17th-18th century, where everyone was fighting everyone for power and doctrine. In essence, it tries to find a way of getting humans to co-exist without killing one another. Notwithstanding the historical practice, the market plays a crucial role; it allows for a way of existing without (too much) blood. The failure of the USSR and other big state projects gently vindicates that line of thinking of the market as the least bad of all options.

The second pillar of liberalism is progress – which ties nicely with ideas of self-actualisation. We are meant to be better off today (both in a personal and societal level) than we were yesterday. One cannot possibly miss the strong Christian undertones and the links with Irenaean theodicy. Early Christians were straggling with the problem of evil; if an omni-benevolent divinity created this world, why is it such a mess? The Irenaean theodicy is pretty simple – and resembles a role playing game; you were put in this world to help you forge your soul though adventures. Clement of Alexandria pushed this even further, claiming that you if you do this successfully, the angel that is watching over you ascends as well. Theodicies are a way out – without them we all would turn gnostic and treat this world as irredeemable. Liberalism turns this self-actualisation to a universal principle and adds a mechanism to it. A more modern take of the whole “we can fix the world” debate is clear in some radical texts – quoting from the Xenofeminist manifesto:

“The construction of freedom involves not less but more alienation; alienation is the labour of freedom’s construction”

“XF is vehemently anti-naturalist. Essentialist naturalism reeks of theology — the sooner it is exorcised, the better”

“We ask whether the idiom of ‘gender hacking’ is extensible into a long-range strategy, a strategy for wetware akin to what hacker culture has already done for software”

Beyond the naivety of what hacker culture is, one can clearly see that yearning for salvation through technological developments which might not exist for centuries. Liberalism, in stark contrast, brings universal methods for salvation here and now, through the market; if you are smart and nice enough, you will succeed, whereas if your moral flaws and incompetence get in the way, you probably deserve the punishment the lifetime of suffering the gig economy is ready to hand you.

For anyone with a more progressive bend, access to the market must be fair, so as for salvation to be just. As liberalism moves on, it removes the last vestiges of irrationality (e.g. racism, sexism, prejudice against regional accents and mannerisms) from the market supply, allowing all to take part in the new Jerusalem. Liberalism is aiming for the likes of automation, faux meat and nuclear fusion to help somehow ameliorate the suffering of the losers. The end-game of this process of extreme atomisation, absolute fairness and winner take all is the lottery, where we override the implicit genetic lottery to create one imposed by society – excluding the already rich of course, for whom salvation is a right of birth.

Delusions and prophecies

Following the 2008 financial crisis spectres, vampires, werewolves and other beasts seemed to be haunting Europe, but it looks like none came out to bite. The left social-democratic strategy at the time had been to try to convince the powers that be that there was some eminent collapse due to finances, the environment, pestilence, wars abroad, dissatisfaction at home, divine retribution and the like. There was an endless stream of predictions of the end of neo-liberalism, the end of capitalism and the of this world; well, this end has still to come and there is no indication that any kind of disaster is capable of destabilising the current way of doing things. The strategies of trying to capture major political parties (in the anglosphere) from within failed and any external resistance was quickly vanquished or assimilated; the left is back at square zero. The state, as a collective representative of capitalist interests, will keep removing itself from all provisions of welfare or responsibility towards its citizens, which are deemed as luxuries, and double down on discipline. At this point, it would be prudent for socialists of all flavours to start re-evaluating their position. Changes of the magnitude envisioned from the left might require Leninist uprisings, which in turn require an armed proletariat. It is not clear at all were the conditions for another world war will come from, and most of us in the west have only seen violence in video games. Furthermore, the failures of central planning led to relative underdevelopment in Eastern Europe and made the socialist model unappealing to broad swaths of workers. What if capitalism is nowhere near collapse and meaningful automation is nowhere in sight, but things can only get worse? What if the collapse of the USSR made the princes of this world feel omnipotent – there are no communists with guns any more, so no socialist path. What if, like early Christians, great ghost dance Indians, the Ezekiel prophecies, Jehova’s witnesses (and a host of other eschatologists) socialists are now facing the issue of no eminent end-of-the-world event, no labour altering technological disruption, nothing worth hoping for, but things are to go business as normal for generations, with the casual disaster here and there? Where do we from here?

The problem of evil and cybernetics

Theodicies go hand in hand with prophecies. If salvation prophecies were fulfilled, if Jesus did descend from the heavens in a chariot and created a new world, the whole problem of evil would not exist in the first place. Without material fulfilment one cannot expect much but an eternal night, a gradual slow death, pummelled by entropy without and our own biology within. For a true believer, however, failure can never come. The prophecies would need to be re-interpreted in an idealistic fashion. If we can’t affect the material world, we could potentially change the way we behave. The failure of prophecy fulfilment – the gospels are very clear on Jesus coming back within a generation – brought to early Christians the problem of actually existing without divine help. This in turn made them delve more and more into questions of ethics, adopting positions from other Hellenistic philosophies (e.g. the Christian position on marriage is largely stoic). Writings of the type of “Didache” – or the teaching of the apostoles, the Three Steles of Seth describe respectively how to worship and how to ascend. We cannot change the material, let us change the spiritual.

So how should socialists behave when there is no prospect of material change? Well, we can reconfigure what technologies are already out there to our advantage – build new platforms and rework our communications and control infrastructure. Those platforms should bring us together, not to (solely) protest, but to have fun, discuss, play and build. Give our lives meaning and get us to meet each other. Integrate us back to nature and help us fill our weekends with more than binge drinking, netflix and packaged holidays. Give us an identity of who we are, how our children are meant to be named and raised; what foods we are meant to eat, how and when to cook, exercise, what relationships to pursue and which to leave behind; when to feel shame and when to feel pride. Help us modulate our inner world; prescribe the right hallucinogenic doses. We are all too secular to worship anything, but smart enough to realise we need sensible defaults to fall back to. Give us a world to be part of – break the cycles of alienation we are subjected to through work, advertising and consumerism. Manuals for a game of life in which we can improve and be in the flow. Provide us with tools to re-examine the causal relationships, turning us all into scientists. In short, technology to create the conditions of a rewarding life, not one of minimum regret.