The Learning Machine, pecking pigeons and the Sending of Being.

Learning is abandoning us.

Back when I was a philosophy student, one of the most talked-about books was Douglas Hofstadter’s The Mind’s I – fantasies and reflections on Self & Soul. First published in 1981 it consists of a set of thought-experiments that tend to make you unsure of some of your most fundamental beliefs.

One of the chapters is called ‘A Conversation with Einstein’s Brain’ and you can read it for free here although I should warn you it is written in the long, reflective form that has now become obsolete. Put simply, the argument is that if you could scan every cell and connection in someone’s brain before their death, you could continue a conversation with them long after their demise – it would just be a matter of calculating inputs and outputs.

In the essay, Hofstadter reasons that you could put Einstein’s brain into a (big) book, and by turning the pages continue to talk to him. As a philosophy undergrad I recall debating how you could substitute anything – pecking pigeons, for example. It had something to do with VonNeumann machines and Alan Turing.

Strange as this may seem, it’s relatively obvious (unless you have a ‘spooky stuff’ view of the mind.): your mind is an emergent property of your brain, your brain is made up of simple cells whose relevant functions could be duplicated any number of ways. It’s a practical, not theoretical problem.

So imagine you were going to build a computer to translate English to Spanish from pecking pigeons. You can picture the various approaches along a spectrum: at one end, you could choose only one pigeon and train it to do all the things you wanted your computer to do – but pigeons have their limitations and are not very easily trained. At the other end you could have millions of pigeons, each doing very simple tasks. What determines the optimal strategy? It all depends how well you can link each processing unit to one another. If you can’t link them at all –  well then you are stuck with one pigeon. If you can link them very well, then it is best for each pigeon to do only simple tasks that they can perform very quickly. If it’s somewhere in the middle, then you’re going to be doing some laborious pigeon-training. As the connectivity improves, the nodal processing simplifies in an inverse fashion – and the overall computational power improves.

From the perspective of learning, humans are disappointing hosts - barely more than monkeys. Their distinct advantage comes from culture - from the invention of writing and books that allow them to compensate to some extent for their pitiful ability to store and process information by externalizing it. But still, they link one book to another inefficiently – remembering the odd bit here, linking it to the odd bit there. And a huge amount of learning on their part is required just to get them to the point where they can read and write. They pride themselves on their learning ability, but really they are only good at processing tasks related to the ‘4Fs’ (feeding, fighting, fleeing & reproduction) – everything else is a bit of a grind. More complicated things take twenty years or more. As information becomes more dense we require more specialism - and even longer to reach the point of progress. Over the centuries, culture became the slow computation.

But…

If you could connect people together more efficiently you could build a better machine – a machine in which people perform simpler tasks - tasks to which they are better suited. Visual processing tasks, for example. They wouldn’t have to learn much at all – and in return the machine’s learning & processing ability would grow exponentially.

So there is a reason why technology is changing the way in which people process information, removing learning and deep reflection and substituting more visual, superficial processing. And it is remarkable how rapidly this repurposing is happening. It is affecting people of all ages, not merely the ‘digital natives’. The essential processing functions of billions of people have been changed; they sift through images on Facebook and click on those they like.

People struggle to imagine purpose without person. Although we frequently stumble upon these bigger, emergent, purposes - evolution, religion, capital, technology – we don’t like to think of ourselves as their component parts. Because we glimpse but not grasp them, we like to think of them as products of human activity – even the idea that a Zuckerberg or Gates is in charge is preferable. But in reality we are like the ant who thinks ‘I wonder what all those ants around me are doing?’

Heidegger understood this: he realized that at best we glimpse these higher purposes, and that we experience them as an unfolding, as uncanny – as a revealing, a ‘destining’ and a ‘sending of Being’. We sense that something is being done to us, but we can’t guess what until we see it.  The nadir of philosophical thought is the realisation that there is something being ‘sent’ in the unfolding of destiny.

People have shifted extraordinarily rapidly from a world of learning, to a world of referencing. The next step – Google Now – will remove the need to look things up, directions will simply appear – much like SatNav. It will still feel as though we are autonomous, but our behavior will form part of a larger computation – a computation articulated through capital and culture, technology and desire.

In the midst of this, Learning itself is ascending upwards - withdrawing from the monkey-heads and libraries and thriving in its new technological strata. Perhaps some of those of us working in learning realize this – that we are either left behind or instruments affecting this change.

There is only one sticking point; learning still needs the monkey-heads to produce the technology. But the end of that era is fast approaching. We are the bridge between one era and the next.