Speaking at the CIPD Ace conference in Manchester yesterday (8 November), Susskind said there was a misconception that only blue-collar work would be automated.
He said: “Many are comfortable with the fact that technology will have a disruptive impact on what blue-collar workers do, but less comfortable with how it will disrupt what white-collar workers do.
“This is because we think blue-collar work is routine and easy to explain but white-collar work requires creativity, judgement and empathy which we don’t think can be automated."
Read more: AI will end all jobs, Elon Musk says
Susskind, however, said there were many examples of white-collar work being automated.
In education, he said, more people had signed up to Harvard online courses in one year than the total number who had attended the university in its entire history.
Meanwhile, in healthcare, Google’s DeepMind has created a system that can diagnose 50 eye problems as accurately as top ophthalmologists.
The AI fallacy
Susskind said that AI achieves these specialisms through a completely different method to humans.
He said: “Often, experts in their field can’t explain why they are so good at something.
“For example, if you ask Gary Kasparov [the chess expert beaten by AI system Deep Blue in 1997] why he is so good at chess, he could give you a few opening moves, but he’d end up saying it requires judgement, gut feeling and creativity.
“Meanwhile, Deep Blue [the AI that beat him], didn’t have creativity or gut feeling. But, it was able to calculate 330 million moves a second, where Kasparov was at best able to hold 110 moves in his head. So the AI wasn’t replicating Kasparov’s skill in order to play, it was playing in a fundamentally different way.”
He described this as the ‘artificial intelligence fallacy’, meaning we assume the only way to perform a tasks at the same level as a human could, would be to do it how humans would do it.
Susskind added: “AI will be capable of performing tasks that requires subtle factors like creativity, judgement and empathy, but performing them in fundamentally different ways.
“AI doesn’t think, it isn’t conscious, but that doesn't actually matter because it can use processing power to perform these tasks differently. This means white-collar workers should be taking the development of AI very seriously.”
Susskind's talk recalled the words of Patrick Winston, one of the earliest AI researchers: “There are lots of ways of being smart that aren't smart like us.”
Ethics of AI decision making
Even if AI is able to handle complex, creative and judgement-based roles, should it?
Putting life-changing decisions in the hands of AI will surely raise ethical objections.
Susskind said AI cannot handle judgement calls better than humans.
He said: “We shouldn’t be asking if a machine can exercise judgement. We should be asking: ‘when do we ask human experts for their judgement on something?’
“The answer to that is ‘when we’re uncertain’, when we have unclear facts, or when we’re in an ambiguous situation.
“The second question is if AI can deal better with uncertainty than a human can. And in many cases, of course it can.
“Take the example of an AI model that can diagnose when a freckle is cancerous. The AI has a database of 140,000 cases to recognise a pattern in, far more cases than a human dermatologist can hope to see in a lifetime.”
Training for an automated future
Susskind said employers and employees of all kinds are facing mass redeployment and urged them to respond through education and upskilling.
He said: “What seems very likely is an increase in demand for human beings to do work fully composed of things that are hard to automate, but currently we spend a lot of time training humans in things like document review and assembly, which AI can be used for.
“With such a great amount of uncertainty around the future, our best response is flexibility. We need to be willing to learn and engage with education single-mindedly.”