Dr Fabian Stephany
Departmental Research Lecturer
Fabian is a Departmental Research Lecturer in AI & Work at the Oxford Internet Institute.
“History does not repeat itself but maybe it rhymes.” Mark Twain (attrib.)
The past of work… is a story told again and again
Today, we are witnessing a technological revolution, as the capabilities of Artificial Intelligence (AI) mesmerise the public on social media and the news. While the power of large language models, like ChatGPT, or generative AI, such as Dall-E, is almost “indistinguishable from magic”—echoing the third of Arthur C. Clarkes’ “three laws”—the public outcry on how this new technology will lead to mass unemployment is not novel at all. As printing machines, cotton gins and steam-powered looms started to rattle away in the eighteenth century; as assembly lines and manufacturing robots were set in motion in the twentieth—and as Siri and ChatGPT have begun to baffle us more recently with their infinite wisdom—a similar narrative has unfolded again and again: A new machine shakes our world of work, every revolution more potent and terrifying than the last, and warnings are heard about how this invention will be our last. The final nail in the coffin of human labour. And yet—while weaving looms, robots, and AI are supposedly destroying jobs—the world has seen an unbroken demand for human labour.
The future of work… remains unpredictable (whether we like it or not)
Obviously, dystopian commonplaces are not the right place to look for reliable answers. And perhaps we should try to resist the temptation of “gazing deeply into the crystal ball” altogether, as most judgements about the future of work have not aged well. In fact, in our recent paper [1], we replicate several famous automation studies and show that estimates range from one tenth (9%) to more than half of the population (54%) of the workforce as being at risk of technological unemployment. But astonishingly, these estimates rely on exactly the same expert opinions. The only thing that changed was the statistical models used to calculate automation risks for the overall population (see Figure), showing that even with the same data at hand, the future of work remains unpredictable, whether we like it or not.
Figure 1: We replicated renowned automation studies and show that estimates for the workforce at risk of technological unemployment vary significantly. While all the studies rely on the same expert opinions, they use different models to calculate automation risks for the overall workforce. For further information see Lorenz et. al (2023) [1].
The present of work… is polarised and craving for human labour
While dystopian imaginaries about technological mass unemployment easily catch our attention, it might be more helpful to map how human work is already affected by AI. Surprisingly, today’s AI-struck labour market is craving for human work but of very different types. Here are two examples. On the one hand, the use of AI has led to the emergence of what the Harvard Business Review has called the “sexiest job of the 21st century”, the prestigious occupation of the Data Scientist. “As companies wrestle with unprecedented volumes and types of information, demand for these experts has raced well ahead of supply”—and so have their wages. On the other hand, AI also gave birth to one of today’s most precarious jobs: the Clickworker. A person who sorts through vast oceans of data in order to clean samples of images and text to train AI applications, including ChatGPT—often encountering disturbingly violent, hateful, or pornographic content along the way. It is a job that is essential to ensure the seemingly flawless functioning of AI algorithms but which is often out of our sight, being frequently outsourced to the “digital sweatshops” of the Global South.
Upskilling: The Challenge of the Century
The examples of the Data Scientist and the Clickworker are in fact two sides of the same coin. For some occupations, the demand for human labour is skyrocketing; not despite, but because of AI. Most likely, AI will both destroy and create occupations. However, the new jobs it creates will require different skills to the old ones—technological change is not skill neutral, as economists say. And so it comes as little surprise that more than half of European enterprises find it hard to fill their vacancies for specialists in ICT. To prepare workers for the millions of new jobs that AI has already started to create, and to cushion those who will lose their work to AI, organisations will have to provide significant resources for the up-skilling and re-skilling of their workforces [2].
In our current research project (www.skillscale.org) at the Oxford Internet Institute, we show how online generated data can help workers and employers understand the skill requirements of new jobs. This approach can offer targeted and near-real time skilling advice to workers, on both industry needs and the skills required to meet them. It also supports policy proposals for recommendations on individual learning strategies and more flexible certification of competences developed through vocational training, short courses or training programmes—so-called “micro-credentials”—which are desperately needed now, in order to ensure that workers and employers are ready to meet the future of work—whatever it may look like.
Dr. Fabian Stephany is a Departmental Research Lecturer in AI & Work at the Oxford Internet Institute (OII), University of Oxford, where he investigates the emergence of new skills and sustainability of novel occupations in times of technological disruption.
Further reading
[1] Lorenz, H., Stephany, F. & Kluge, J. The future of employment revisited: how model selection affects digitization risks. Empirica (2023).
[2] Stephany, F. and R. Luckin (2022) Is the workforce ready for the jobs of the future? Data-informed skills and training foresight, Working Paper 07/2022, Bruegel.