On the mechanisation of thought

The industrial revolution mechanised physical labour. We went from using machines to grind flour to assemble cars and then some in less than 200 years that saw unprecedented innovation. The holy grail of machanisation remains however: thought.

As mechanisation relieves workers from manual labour, the work-force tends to move towards more sophisticated work. Sophisticated work takes time to learn; since the industrial revolution, the age at which we start giving back to the economy has increased steadily. The natural conclusion of this needs careful consideration. If I’m right – that is, if this trend continues – people’s lives will fundamentally change when we lose the competitive advantage we have always had over machines, our intelligence.

The breakpoint is when there is no higher level of sophistication that the work-force can competitively move to – when (almost) all jobs that produce things that people are willing to pay for are performed better or cheaper by machine/computer/robot than person. Note that we’ll get to this point long before we create strong AI. For those concerned about jobs – and in this society that should be most of us – this is a huge deal. What jobs will still be done by people? How will resources be (re)distributed? What role will politics play? How will society transition to a new steady state? This should be the topic for economists and political thinkers today, but there’s almost no discussion of it.

I can think of two reasons why this is not widely debated: people think it won’t happen in their lifetime or people want to avoid thinking about something that is likely to cause them worry.

The latter reason is likely more common. Ignoring a problem doesn’t make it disappear though. If people did think about it they would realise mass-unemployment (coupled with increased productivity) is likely to happen in our lifetime. It might be hard (remember the AI optimism in the 80’s), but there are no fundamental barriers to mechanise much of what we are willing to pay for with faux-intelligent computers within decades; in the end, the brain is a kind of computer.

We can argue about the timeline. I think serious arguments putting it more than 100 years into the future will be rare; I would argue for it to take less than half a century and probably significantly earlier. The key is that innovation increases the sophistication of the tools we use to push innovation further, making the trend exponential (it’s not quite that simple, but roughly right). Timing, however, is not the point. The point is that we’ll need to deal with it, bar a catastrophic event.

The solution is in theory simple: redistribution in the form of a citizen salary. And I say that despite libertarian leanings. Politically this shouldn’t be too hard to argue – widely established practices like unemployment benefits, social security, and state-funded healthcare already gives something for nothing.