The march of unfettered automation can be slowed, but tech companies will not do it by themselves.
There were two directions in which artificial intelligence (AI) could have gone. The first is the development of autonomous machine intelligence of the type that would pass Alan Turing’s test and imitate a human so well that its responses to questions are indistinguishable from a human’s. The second is ‘pro-human AI’, the creation of intelligent machines that we can work with and that are helpful to us. GPS, hyperlinks, and many aspects of the internet itself are examples of that ‘machine usefulness’ – although even they need to be managed to avoid unintended consequences.
The first route produces AI that can replace human labour, destroying jobs and lowering wages. The second enhances human capabilities, creating more value and new products and services.
There are no prizes for guessing the route that we are now progressing down, further and faster than ever.
This panel event, welcoming Professor Daron Acemoglu as the Sanjaya Lall Visiting Professor of Business and Development at the University of Oxford, explores how we can change direction and navigate the narrow AI path that will be good for workers and humans.
Professor Acemoglu is joined by Professor Sir Christopher Pissarides and Professor Helen Margetts in a discussion chaired by Professor Sir John Vickers.
The speakers consider the ‘roadblocks’ that stand in the way of a positive and pro-human AI: excessive automation, loss of information diversity, misalignment between human cognition and AI algorithms, and monopolised control of information. They point out where a rush to replace humans has not worked well – the ‘so-so automation’ of supermarket self-checkouts for example. And they warn that the monetising of data practised by the big tech companies will lead only to more ‘filter bubbles’ and misleading or extremist viral information. But they also show where AI tools can have unexpected benefits: for example, ChatGPT can help lower-performing workers with writing tasks, and in the public sector Big Data can improve targeting of policies.
