Data or decision – two responses to AI

(Dette er et re-post fra LinkedIn for at have essayet her på bloggen også)

There are two fundamentally different responses we, as a society, can have to robotics and AI. We should work very hard to choose one over the other.

The fundamental characteristic of AI, robotics and the resulting latest wave of automation is that there is no data processing that will be out of reach of algorithms because of complexity in a few scant decades. This would be no real problem, if it wasn’t for the fact that almost everything that the market values can be described as – and therefore conceived of and programmed as – data processing.

When I take knowledge of a subject and shape it in order to transmit it effectively to students, I am taking one set of data – knowledge – and cross-referencing it with a lot of other data – the time-frame of the course, the make-up of the student body, reacting to complex facial and verbal cues in teaching situations, and so on and so on – and delivering the results of this rather complex data processing.

At the moment, there is very little doubt that the most effective data processor for this kind of work is the human mind, since so much of the necessary processing occurs automatically.

This is what’s happening when we find that we have a talent for something. Teaching comes relatively easily to me, because a lot of data processing is going on in the background, while many teachers – who do a very good job, I hasten to add – must explicitly formulate a lot of the data processing for it to work. They don’t have the gift, the feel, the talent for teaching, we say. But with the right amount of processing power and the right programming, an artificial intelligence can be programmed with that gift.

It's not just me

What can be said for my current main profession can quite easily be said for whatever else I do for a salary – evaluating proposals, analysing strategic challenges, writing non-fiction books – and what’s more chilling, is that it can be said for almost every other profession there is.

Most of the work a lawyer does – checking precedent, following procedures, constructing arguments, even picking juries – is data processing. Most of what a doctor does – diagnosing, treating, reacting to the effect of medicine, writing correct prescriptions – is data processing. What a successful financier does – picking stocks, knowing when to sell, optimizing revenue for an IPO – is definitely data processing, and there is nothing the market values more.

You can’t escape in the arts. Look at what the market values in movies, for example: tightly structured plot, consistent characters, clear visual language – all of that is data processing. When the internet explodes in criticism of the plot holes of a new film, the criticism is shaped as a criticism of the inferior data processing of the moviemakers. Our language for understanding and appreciating – valuing – art is purely focused on data processing.

You can’t escape in politics. What was hopefully described by Hannah Arendt as the true existential arena, where bold and decisive action of enlightened individuals shaped policy and decisions, is now usually conceived of as data processing. This is true whether you are a populist – you must interpret the data of the will of the people – a liberal/centrist – you must construct objectively superior solutions through utilitarian processing of societal statistics and data – or even a follower of Donald Trump’s marketing approach to politics – you must process all data in a way that can back your own narrative. Trump’s victory is – many experts say – simply a victory of superior data processing by his sub-contractors.

But surely, interpersonal relations must be outside this realm? Well, are they really?

There is no shortage of courses in body language, where we learn to read data points from the facial and bodily movements of our colleagues. We conceive of the mind increasingly as directly the same as the various neurons firing in the brain at any given moment, attributing our deepest feelings and experiences to data processing. For the moment, these two data processing tasks are outside the full reach of algorithms. Soon, though, professions that rely on interpersonal relations – for many pundits the safe haven for a future job market – will be out-performed as consistently by robots and AI as cancer diagnosticians are now.

The market rewards data processing

The market in every field rewards the superior data processor. To a certain degree, this is because we culturally value this, and the market to a degree facilitates and strengthens what we value culturally.

But it is also because the market in and of itself is a gigantic data processing mechanism. It works only with free flow of data – our use of currency – and the output of the market is exactly the output of data processing: a general response to a wide range of individual outputs.

Many individuals make individual decisions about what they value – they attempt to buy what they want – and the market will then provide products that are as close a response to that demand as is feasible.

(This, by the way, is at the core of why the market is not a very good engine for creatingpersonal freedoms – the market cannot effectively generate personal, individual outcomes, only general ones. You may vote with your wallet, but only a million wallets will constitute a potential customer segment.)

Can data solve data?

So what can we do?

Well, from post-New Labour social democrats and all the way to the populist right, the answer is clear: the market will provide. This is their basic ideological response to everything, so it’s not a surprising answer. And it might work, if it wasn’t for one thing: it’s too slow.

It’s a pretty good bet that the vast technological evolution I outline above will come to pass inside the next 20-30 years or so. If the market should provide solutions, we must as a culture change what we value – away from superior data processing to something else – before this time. That is not something that seems feasible, since almost all our conceptions of quality are centred on superior data processing.

What the market if left unchecked will provide is not new, lovely jobs, but instead a lot of jobs that can pay only so much that a robot can’t do it cheaper. The human work force will only exist as a cheaper alternative to the robot work force.

This is not an attractive future, and it is frankly not a very sustainable future. In this scenario, we revert from an economy that provides some – limited – economic and social mobility through increased income equality, to an economy that is only based on existing fortunes and wealth. This is how economies were before the industrial revolution, of course, but back then the wealthy had ample reason to want to preserve a lot of the less fortunate: their wealth was built on the labour of the lower classes.

While I - perhaps wrongly - hesitate to ascribe directly malevolent ideas to the rich, it is an open question how long the free market flow of their wealth through investments will flow towards unprofitable humans, when there are profitable robots they could invest in. Of course, the benevolent billionaires that already exist today would want to preserve at least some of the lower classes as well. But is this a politically desirable outcome for humanity: that the bulk of us will exist as pets for the few?

Can we make a decision to change?

The other way to go is to regulate the market in some way to compensate for increased automation. This can be done in multiple ways, one of which is universal basic income. The idea that every person gets a living wage from the state coffers each month.

This would move the basic value we, as a society put on people, out of the hands of the market and into the hands of the individual, who would no longer rely on other people’s appraisal of her value to survive. Incidentally, it would also completely obliterate the absolute necessity for job creation in an increasingly automated economy.

It’s a way to try to force the cultural change that would need to happen before the market could take care of the problem itself. It is a way to attempt to change the conception of a successful life away from someone who is a great artist, superior salesman, clever financier, insightful doctor and so on into: someone who likes what they’re doing.

There’s plenty of problems with universal basic income – how to handle debt and mortgages is a particularly pressing issue in many national economies – but it does have one quite clear advantage over the alternative answer of letting the market decide:

It is not automatic, but an actual action.

It is a response to the situation that is constructed not because it arises from what has gone before – data – but from what we want to build – decision.

So, I don’t particularly want to advocate universal basic income as the only possible response to the challenge of robotics and AI. What I want to advocate is that it is the proper kind of response to this challenge. It is the kind of response that takes control of our common future away from automation and generalisation and into the imaginations and dreams of individual, unique human beings.

It is the kind of response we owe to ourselves, to our fellow human beings, and to the hopefully thousands of generations of human beings to come.