Yes, AI will wipe out some jobs. But this is the story of progress
AI will inevitably make some jobs redundant, but the benefits to society are far worth it, writes Paul Ormerod
The world of AI has gone through exciting times in the last couple of weeks.
The main event of course was the launch of the Chinese AI app DeepSeek. It looks very much like ChatGPT and almost immediately shot to the top of Apple Store’s downloads.
There are still some important questions around DeepSeek which remain unanswered. But it was developed at a fraction of the cost of ChatGPT.
The developer of ChatGPT, Open AI, launched a different product only this week. If the claims around it stand up to robust scrutiny, it could prove to be at least as significant as DeepSeek.
The new tool, “deep research”, put together reports which OpenAI says match the output of a research analyst. The company claims its tool is a significant step towards the goal of developing artificial general intelligence.
How many jobs may be lost to AI?
One of the fundamental questions around AI in general is what its impact on jobs will turn out to be. By nice coincidence, a paper has just been published in the high ranking Journal of Economic Behaviour and Organization on the impact of ChatGPT on the demand for freelance workers.
In essence, tools like ChatGPT are simply another form of automation. They enable tasks to be performed by capital – a machine, an assembly line – which were previously carried out by labour.
The substitution of labour by capital has been a consistent and powerful theme since the start of the Industrial Revolution in the late 18th century. The economic system under which we live is, after all, known as “capitalism”.
We do know the answer on what the effect is, at least as far as the past is concerned. The workers whose jobs are directly impacted lose out.
But the increase in productivity which capital investment brings benefits everyone else. The relevant product becomes cheaper and usually of better quality. The consumers benefit, if only by a small amount. But there are many more consumers than the workers who make the product. So overall, society benefits.
The fear around AI, certainly in the public perception, is that it is potentially such a pervasive technology that it might create large numbers of workers who lose out from its application, swamping the gains made by consumers.
The issue has been the focus of intensive research in the past few years. The overall consensus is that the worst fears around AI are unlikely to be realised. For example, a well-publicised study in 2017 claimed that at least half of the US labour force could be in jobs which were at risk of automation. But subsequent refinements of the analysis suggest that less than 10 per cent of jobs are at high risk.
This is the story of progress
The study on the impact of ChatGPT on freelancers fits in neatly with the emerging view that AI will not lead to an apocalypse in the labour market.
The paper uses advanced technology to analyse 3m job postings on relevant online labour markets. Such postings are highly flexible and quick to react to changes. The authors identify 71 very finely grained skill clusters which they categorise as being either substitutable by or complementary to AI.
Following the introduction of ChatGPT, the results do indeed show a reduction in demand of approximately 24 per cent for skills deemed substitutable by the new technology compared to unaffected skill clusters. But complementary skill sets such as chatbot development and machine learning showed a distinct increase in demand. And overall a greater proportion of skill clusters are complemented by AI rather than substituted.
It is still early days, but the evidence suggests that the ChatGPT technology will yield substantial benefits to society as a whole.
Paul Ormerod is an honorary professor at the Alliance Business School at the University of Manchester and an economist at Volterra Partners LLP