GRAHAM BARR AND BRIAN KANTOR: How AI will affect the workforce
ChatGPT is the latest incarnation of artificial intelligence. The technology behind it is something of a marvel, but it may entrench labour inequalities — globally and locally
There’s no dispute that ChatGPT has captured the imagination of society. The idea that you can type “write me a three-page essay on the causes of the French Revolution” and get back an articulate paper, at the precise length, is extraordinary.
An exemplar of the inexorable rise of computer technology, it has received particular prominence because it is so easy to understand. The model looks for consistent patterns and features of data to “train” on, and then produces an output based on this. Generative artificial intelligence (AI) of this sort is rightly considered something of a marvel. But it will have a multifaceted effect on the world of work.
Consider interactive chatbots. Based on parallel technology to ChatGPT, they’re already used — if not particularly adeptly — to answer client queries in the banking space. That technology is part of the move towards internet-based, teller-free banking, as are loyalty programmes that reward online and app-based transactions.
The bots may be in their infancy, but as they improve they will undoubtedly become entrenched in those service industries that have traditionally relied on call centres to interact with clients — think banking, insurance and telecommunication companies — taking a toll on this semi-skilled workforce. But AI-driven chatbots could also extend to areas such as tailored investment, legal or even medical advice. Affecting the more highly skilled, in other words.
There is a flipside: if legal, financial and medical advice become easier to give — and less expensive to deliver — the real price of such advice will fall and demand will increase. Clearly beneficial to the consumer, it will in turn lead to increased employment for those who can still add value to AI, relieved of drudge work.
As in the services sector, AI will drive efficiencies and profitability in retail. The huge data sets discoverable from loyalty programmes, for example, will allow accurate targeting of products and stock management. That will eat into midlevel management in retail groups.
Taken together, AI will affect a host of semi-skilled jobs in the retail and services sectors; bank tellers, call centre operators, insurance agents and retail store cashiers, for example, will become redundant.
There are, of course, those critical areas to which AI cannot contribute. Computer software has a much smaller role in directly productive areas such as construction and law enforcement, or primarily low-skilled, labour-intensive activities such as agriculture. It can support robotic mechanisms that aid production, but it cannot build houses and bridges. It cannot directly provide security. And its application to sowing seeds and harvesting particular crops is limited.
Still, even within these job categories AI will encroach more on skilled and semi-skilled jobs. In agriculture, for example, drones with AI will identify plant diseases from photographs and suggest appropriate remedies. Drones can monitor and provide photographic coverage for security purposes, and AI can assist criminal prosecution by rapidly analysing databases and invoking facial-recognition software.
Within the care economy, jobs involving people-centric skills will not be greatly affected. Child care, nursing, counselling and care for the elderly are obvious examples; managing sick and injured people in hospital is also not a clear avenue for AI applications. But jobs that are regarded as highly skilled will be affected: the skills of doctors and specialists to diagnose diseases, for example, are already quickly being taken over by AI. Delicate surgery empowered by AI is already prevalent.
In short, when skills are human-centred or require an empathetic touch, they are still not replaceable by AI. But when human skills are learnt and technical — however complex — they are likely to be easily displaced.
Any sparkling computer science talent that emerges from universities in developing countries such as South Africa will be in the forefront of the emigration exodus
AI in the world
What does this mean for a world of AI? What could it look like?
For a start, society-wide per capita wealth in computer-centric countries such as the US will continue to rise. With that, any sparkling computer science talent that emerges from universities in developing countries such as South Africa will be in the forefront of the emigration exodus.
At the same time, those in the developing world who lack local work opportunities but have the means to leave will continue migrating to developed countries to offer services — often in the human-centred but less skilled areas AI cannot affect, such as care of the elderly.
South Africa will have a typically quixotic response to AI. It will feed off the technological advancement in the hope that its efficiencies will allow for growth. But AI is no panacea for the local economy — certainly not on the all-important question of rising unemployment. In the absence of any rapid turnaround in the failing education system, the country will likely remain, for the most part, a resource-based, extractive economy where increasing mechanisation and computerisation will lead to a gradual decline in employment.
In many service-orientated sectors, such as banking and insurance, AI will become increasingly entrenched and lead to rising profitability but falling employment. Successful high-employment industrial sectors, such as the Eastern Cape-centred motor assembly plants, will become increasingly robotic and shed jobs. Yet other labour-intensive (if people-centric) islands, such as tourism, will flourish.
Overall, one envisages a continuing low-growth, low-employment economy constrained by a dysfunctional government — an economy that continues to be at the mercy of the commodity cycle and a weak educational system. All of this will result in rising inequality and an increasing proportion of the population depending on welfare.
* Barr is emeritus professor of economics and statistical sciences at UCT; Kantor is emeritus professor of economics at UCT
Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.
Please read our Comment Policy before commenting.