A recent report by the Royal Society into machine learning highlighted the importance of good PR when it comes to AI and the various risks and benefits associated with it. In my opinion, it calls not only for the industry to talk wisely on the topic, but also on the press to do their job and properly research stories.
A good example came recently via the BBC, where an article was headlined by “Automation poses a high risk to 1.2m Scottish jobs, report says”. The report in question was a paper released by the IPPR that explored the impact of technology on jobs in Scotland, and in turn on the changing skills landscape we are likely to see in the future.
This paper was itself a more focused version of a similar paper published earlier in the year that explored the same issues from a UK wide perspective.
My problems with this are two-fold. Firstly, the automation element of the report itself is based wholly on the famous 2014 study from Oxford University on the risk of automation to jobs. The paper received a huge amount of publicity as it painted a dark picture of the future.
Unfortunately, the paper was highly flawed for two main reasons. Firstly it made a huge logical leap, as their analysis looked at the potential of any part of a job being automated, and then extrapolated from that that the entire job is liable for automation. It’s akin to suggesting that writers will be automated because spell-checkers automate spelling for us. Indeed, it’s something that Carl-Benedikt Frey himself admits today.
“In many ways, robots could enhance careers rather than destroy them. The introduction of automation in the workplace will usher in a time where our jobs will become more creative and involve more social interaction. Although robots will render some occupations obsolete, as technology has in the past, humans and robots will also complement each other in many tasks, creating new types of jobs,” he says.
The study also succumbed to the lump of labor fallacy, and didn’t suggest any new jobs will be created whatsoever. Technological disruption is inevitable and has been with any new technology, but recent studies have found that whilst technology may have ‘destroyed’ a few hundred thousand jobs in the past decade, it has created a few million new ones.
Focusing on the negative
So I have grave reservations about the basis upon which the analysis of automation was made. But there is also the reporting of the paper itself, with journalists too often focusing excessively on negative, scary headlines because they know full-well that fear sells.
The IPPR focused relatively little on the threat of automation, and considerably more on the need for skills training and support to be extended into adulthood. It’s a topic that was also covered in a paper from the Science & Technology Select Committee last autumn.
Both it and the IPPR paper make sensible recommendations, especially around mid-career learning. For instance, “A new mid-career learning route, with a mix of online and face-to-face provision delivered through existing providers, in a fully flexible, transferable and modular approach. This route would be focussed on delivering improved rates of career progression, pay and productivity, starting in low-skill sectors.”
Suffice to say, this isn’t to suggest that automation won’t present many challenges to society, but we need balanced coverage to ensure that we don’t overlook the tremendous gains such technologies will bring. It requires an honest and frank dialogue between the public and industry to ensure that both sides of the equation are accurately reported.
As the Royal Society report states, we need “continued engagement between machine learning researchers and the public. Government could support this through a public engagement framework.”