Ever since Oxford University researchers Carl Benedikt Frey and Michael A. Osborne published their report into the number of jobs that are susceptible to automation back in 2013 there has been no shortage of doom-laden prognoses on the state of employment as AI expands it’s reach.
Sadly the paper seems to have inspired a number of predictions that have at their heart a fixed pie approach to jobs, and whilst I’m sure that jobs will be lost to automation, I’m also sure that new jobs will be created because of it.
The seemingly unbalanced nature of the debate thus far makes the recent briefing note from the Parliamentary Office of Science and Technology interesting. The Parliamentary Office of Science and Technology exist to provide UK parliamentarians (and us members of the public) with a balanced and impartial overview of pertinent science and technology issues.
Automation and the workforce
As such, the paper provides a reasonable overview of the state of play with automation, and it’s possible implications, for a relatively lay audience.
It looks at the possible implications for both knowledge work, such as e-discovery and algorithmic trading, and physical work such as driverless cars and automated factories.
Despite the significant potential for improving performance, the paper argues that the key driver behind the takeup of automation technologies is a cost one.
“The pace of adoption of RAS is likely to depend on factors including the cost of technology and labour, awareness of the technology, public perception, and regulation,” they say.
A key aspect of this has been the increasingly cost-effective ways of collecting data, and of then making sense of it using machine learning. With low-cost sensors adding a growing volume of ‘real world’ data, this is only set to continue, whilst labour costs are largely going in the other direction.
The impact on employment
The paper goes on to acknowledge that past technological innovations have had a limited impact upon employment levels, and there is scant evidence to suggest that anything different will occur with automated technologies. What studies have been done fail to account for the new jobs that will be created, and thus succumb to the fixed pie fallacy.
It goes on to highlight the rapid growth in demand for people with skills in machine learning and other AI related disciplines. With tech companies fighting each other for the best talent, this is only likely to continue.
What is more interesting however is the section on how communities and individuals can adapt should their livelihood be automated. We’ve already had a proposal from the UK Labour Party for a National Education Service to mimic the National Health Service in providing free lifelong learning to citizens.
The UK already provides free education to citizens up to the age of 18, but this proposal reflects the growing need for re-training and lifelong learning. Of course, many of the possibilities mentioned in the POST paper, such as MOOCs, are free already so should not incur great cost.
They do however remind us that most studies into MOOC users to date suggest that the student body consists primarily of those who already have degree level education, and thus relying on this method may actually deepen inequality, which is something highlighted by the Matthew Effect.
There may also be significant geographical implications, with some regions more likely to be effected than others. Of course, this has always been the case as industries ebb and flow, but given the struggles many post-industrial communities have faced, it is perhaps a message that warrants repeating.
In such a rapidly changing environment, I’ve no doubt that things will evolve sufficiently to warrant an update of this paper before too long, but in the mean time it does provide a decent overview of the current state of AI, and it’s potential implications on society.