The limits of generative AI mean the technology is not likely to displace jobs in a widespread manner, according to research from two Oxford University artificial intelligence and automation experts, revising predictions made by the pair 10 years ago.
The paper, Generative AI and the future of work: a reappraisal, said AI would instead likely impact wage differentials, productivity and strike action.
The research was conducted by University of Oxford professors Carl Benedikt Frey and Michael Osborne, who co-authored a famous 2013 paper – The future of employment: how susceptible are jobs to computerisation? – that stated that almost half of all jobs were at risk of automation. It suggested jobs associated with low wages and low educational attainment, such as telemarketing and insurance underwriting, were most at risk.
But now the researchers have suggested that the upper limits of popular AI technology’s abilities have been almost reached. Issues with perception, creativity, the novelty of output, social ability and hallucinations (stating untruths with confidence) mean that, while some jobs will be automated, there are still many scenarios where human input is vital. For example, the effective use of robots for warehouse work is ubiquitous. But in higher-stakes work, such as driving, human perception is still critical. While Amazon is automating sales representatives and account holders for more transactional accounts, for larger accounts that require careful relationship management, a human representative is still used.
There are also areas where increased use of AI has created jobs, such as prompt engineers, with the need for a human worker to select, oversee and refine the output of AI only emphasised.
Ultimately, AI is still only able to riff on that which has already been created by humans, Frey and Osborne observed. “For now, deployment of generative AI will be confined to lower-stakes activities, like customer service or warehouse automation, where engineers can redesign and simplify the environment to enable automation,” they explained. “Longstanding relationships, benefitting from in-person interaction, will remain in the realms of humans.”
By dispensing with routine and repetition tasks, AI will only bring human-to-human interactions more solidly to the fore, agreed Perry Timms, founder of People & Transformational HR. “AI will enhance what we seek and expect from in-person interactions, where we outsource things like calendar slots, invites, note taking, action planning and so on in favour of us just focusing on the conversation to be had in that information-exchanging session,” he said.
That said, the Oxford professors’ latest paper does showcase how AI has helped boost productivity in roles such as copy editing and software development, and lowered the barrier to entry to some content creation professions, which could depress wages. As such, there has been a backlash against the use of AI in certain sectors, with strikes ongoing in Hollywood around the use of AI in scriptwriting.
For Clare Walsh, director of education at the Institute of Analytics (IoA), this means businesses should be aware of how AI impacts humans, where it fits into work and how to implement it ethically. She said: “Legislation is coming fast and we are more thoughtful on what responsible AI looks like but there are, of course, new challenges today as the technologies evolve.
“The success criteria for businesses may not be aligned to a fast and easy turnaround, or, if they are, only certain roles can be performed by generative AI, not the whole project cycle and we [the IoA] do not advocate taking human decision making out of the loop.”
Read the CIPD's resources on AI at work here