There has been a trend among many commentators to predict a future in which AI takes over most jobs, leaving humans to depend on some sort of government-issued universal basic income.
This ignores the fact that many opportunities are being created for people to marshal the AI technology around them to create and serve new products and markets. Or to upskill to meet what is becoming unsatiable demand for people to design and maintain the systems behind AI. Or, for that matter, the security, maintenance, and administration required to keep AI systems and servers up and running.
In recent weeks, the angst has been generated by a recent McKinsey Institute report that suggests 40% of US jobs could be allegedly replaced by AI. While the McKinsey authors sought to identify areas where automation could actually boost job roles, the 40% figure is was stuck across much of the media landscape.
Can we really predict what jobs AI will subsume? It’s likely that even with the best data and predictive capabilities (including AI itself, of course), we simply don’t know what jobs or skills will ultimately be on the line in the months and years to come.
Predicting the precise impact of AI on employment is fraught with uncertainty – and most of the time just plain wrong, write Thomas Davenport and Miguel Paredes, in Harvard Data Science Review. Instead of gloomy predictions, they advocate for a shift in focus toward preparing workers for AI-related changes through skills development and job redesign. The McKinsey team, for its part, says as much.
Davenport and Paredes looked at predictions about job loss related to AI going all the way back to 2017, and found none of them have panned out in the apocalyptic fashion predicted. For example, the World Economic Forum in 2018 issued a prediction of seven million global job losses due to AI by 2022.
“We have studied various analyses of jobs predicted to be automated out of existence by AI, defined broadly as machine learning, deep learning, robotics, gen AI, and so on,” Davenport and Paredes explain. “A common attribute of these predictions is that they are wildly inaccurate.”
Even drilling down to tasks, it’s been too difficult to really understand which ones will be subsumed by AI. “The most responsible approach may be to discontinue making them, or admit that any predictions are highly speculative,” the co-authors strongly recommend.
To illustrate, if all the predicted job losses at the high end, two billion, did occur, that would mean the equivalent of every job across the developed world would be lost to AI. The percentages of potential job losses ranged between 5% to 47%, they said.
The prognosis offered in job-loss estimates is mainly negative. Most forecast some job gains from AI, “although most predicted substantially more losses than gains.”
All predicted some moderate-to-substantial percentage of ‘automatable’ jobs—ranging from 5% to 47%. “There are no good data on how many jobs have been lost or gained because of AI, but there is also no evidence of large-scale job loss, and unemployment rates in the economies most aggressively adopting AI remain relatively low,” they added. “If the organization or industry is growing rapidly, for example, it may not be prepared to eliminate workers even if many of their tasks can be performed by AI.”
Let’s get away from the scary pronouncements, and focus on preparing and developing our workforces for the AI era ahead. And helping people to realize and act on the enormous entrepreneurial opportunities it presents. “Workers could be trained in a digital mindset, in specific AI skills, and in reviewing and overseeing the work of AI to ensure quality,” , Davenport and Paredes urge. “They could also practice the redesign of jobs and processes to effectively incorporate AI into new approaches to work.”