Startup general interest

Industry breakdown of jobs subject to computerisation

By January 24, 2014 2 Comments

I’ve blogged before that 45% of jobs are at risk from automation, and this morning our Senior Designer Peter Main tweeted a breakdown of which industries are most at risk:

Screen Shot 2014-01-24 at 11.47.27

It is easy to think about automation in apocalyptic terms – e.g. all telemarketers will be replaced by computers in the next twenty years, but the way it works in practice is that computers make existing workers more efficient and only a percentage of jobs are automated. In service industries this manifests itself in a massive increase in the number of clients one employee can serve.

The underlying technology drivers are artificial intelligence and robotics, with artificial intelligence having by far the greater impact in the short term. We have two companies in our portfolio that are using artificial intelligence to provide services that were previously only available as one-to-one services from people, and in both cases they are making their services available to a much wider group of people than could access them before, and at a much lower cost (I can’t share names as neither is announced), and I expect us to make more investments in this vein. In other words, I think automation is a big opportunity for startups.

I’m sensitive to the human cost of automation but it’s important to remember that it comes with many benefits in terms of reduced cost and wider access to cool services, and hence is a driver of economic growth. Hence I see automation as part of the creative destruction process which sits at the heart of capitalism. The big question is whether new jobs will be created rapidly enough to keep society in balance, and whether workers can acquire the new skills required to do those jobs effectively. The skills required are likely to be social and creative skills.

The table above is taken from a paper titled The future of employment: How susceptible are jobs to computerisation?. The authors estimate that 47% of jobs are at risk from automation over the next two decades, a number they derived by analysing 702 different industries according to whether their tasks are susceptible to automation. It’s a thorough analysis which takes into account likely further improvements in technology, but if anything I think they underestimate the likely pace of development and hence the number of jobs at risk.

Here’s a few other nuggets from the document that I liked:

  • In 1933 Keynes predicted widespread technological unemployment “due to our discovery of economising the use of labour outrunning the pace at which we can find new uses for labour”
  • In 1589 Queen Elizabeth I refused to grant a patent to William Lee for his stocking frame knitting machine saying “Thou aimest high, Master Lee. Consider thou what the invention could do to my poor subjects. It would assuredly bring them to ruin by depriving them of employment, thus making them beggars.”
  • Much of the improvement in machines’ ability to do non-routine tasks comes from big data, specifically the ability to pattern match against past events in large datasets.
  • Healthcare diagnostics are already being computerised. Oncologists at Memorial Sloan-Kettering Cancer Center are using IBM’s Watson to provide chronic care and cancer treatment diagnostics (Cohn, 2013).
  • Sophisticated algorithms are gradually taking on a number of tasks performed by paralegals, contract and patent lawyers, particularly by scanning thousands of documents to assist in pre-trial research (Markoff, 2011).
  • Improvements in sensing technology is making sensor data one of the most prominent sources of the big data that enables automation.
  • Services like Future Advisor use AI to offer personalised financial advice at lower cost.
  • Estimates by MGI suggest that sophisticated algorithms could substitute for approx 140m knowledge workers worldwide.
  • Perception and manipulation, the main challenges to robotics, are unlikely to be resolved in the next decade or two.
  • Realtime recognition of human emotions remains a challenging problem.