The Future of Work – Part 1

Are we headed towards a future that sees large segments of the human workforce  permanently displaced like their former equine partners?  A future where workers do not migrate from one dying sector of the economy to a new and upcoming sector that needs the displaced surplus labor as has been the case so far since the beginning of the industrial revolution?

The 200-year old history of industrialization and technical innovation has been a very positive one … to date. Generally the jobs that have disappeared, overwhelmingly manual labour in industry and agriculture, have been replaced by better jobs, jobs that pay more, require more brain and less brawn — after all we are nothing like horses, we adapt, we learn, we have large squishy frontal lobes. Is not that horse comparison egregiously insulting?

Two-hundred years is a long time, and the pattern of job loss in one area and employment take up in another as new industries and technologies emerge has been almost invariable. But a pattern is only a pattern until it isn’t — we haven’t had a pandemic in a hundred years and we have one now.  Many of those considering the future of work think that there is a pattern-breaker looming — and it is called artificial intelligence and machine-learning.

But why are AI and ML not like other forms of progress through automation? Futurists isolate at least two fundamental things that make AI very different from a horse-replacing tractor. The first is exponential acceleration.  Since the 1950s when the first integrated circuits were produced computational power has doubled approximately 30 times. This is a stunningly steep ascent in the ability of machines to process information. There is no parallel for this since the invention of writing gave us institutional memory as a species.This acceleration is not slowing and its consequences are unimaginable.  

The second difference is cognitive capability. We don’t have to get all Issac Azimov here and conjure up an image of the friendly android next door; but machines are making decisions, solving problems and they are learning, oh yes they are learning. Consider the ancient game of Go. Apparently when playing the game of GO, there are more moves, more possible configurations of the board than there are atoms in the universe. The champion-level players often can’t even put into words what they do, they seem to intuit their most brilliant stratagems. So this is not like chess — a team of programmers can’t just throw brute computational power at a human GO grand master, the way IBM’s Deep Blue did in its matches against Gary Kasparov in 1997 and beat the human. Yet Google’s Deep Mind Group created the Alpha-Go algorithm which eventually defeated the  reigning world Go champion.  How on earth did this happen?  Well, apparently Alpha-Go learned while it was playing and it never forgets what it’s learned.

It seems clear that machines powered by AI and capable of learning will rapidly climb the job skills ladder. It has been thought for some time that jobs that break down into rote repetitive tasks (like commercial truck driving) will go the way of the horse and plow within a decade or two; but those wearing white collars and in particular professionals with hard-earned credentials and years of training were thought to be on the other side of this machine replacement dividing line. But are they?  Radiologists go through years of advanced training to learn to interpret the different sorts of images that are so crucial to diagnosis and treatment decisions in medicine. What if an advanced AI is stored with every single image of the human anatomy that has ever been taken, is capable of cross-referencing them all and comparing them, learning from and never-forgetting errors and has built in vision scanning that takes images down to the molecular level without breaking an algorithmic sweat or needing quaint additive tech like microscopes? 

What of the human intuition that comes with all that training?  Well in five years Alpha-Go would be held-back in second grade, quantum computing will be advancing and who can really know whether the combination of intuition and intelligence that makes us who we are will matter as much in the world of work? 

We don’t have to extrapolate too much here. Let’s leave science fiction futures off the table for now. But even looking only at rote, repetitive jobs, the labour market implications are massive — in 2019, 5.8 percent of all American jobs were in the trucking sector — an industry worth $700 billion. Does anyone think these jobs will be there in twenty years? Estimates are that jobs like this represent 50% of all current jobs. What if even the relatively immediate future brings us a world in which human work for wages is becoming less and less necessary?

In my follow-up to this piece I want to look at one aspect of this predictable future. How can a market economy and consumption, which is currently fueled by wages paid for work, survive and thrive in a much more jobless world?