The impact of AI on the knowledge economy has been a hot topic for several years now. The conversation feels particularly enlivened by the introduction of GPT, which is evolving and learning at an extraordinary rate and which appears to offer greater potential for radical change than anything we have seen before.

As lawyers we are used to hearing how AI will, in time, totally transform the way we work (and litigators worldwide are already grateful for the role technology assisted review plays in electronic disclosure exercises), but we have not heard much about the role AI might play in the broader judicial sphere.

It was therefore particularly interesting to hear Sir Geoffrey Vos’ recent thoughts on the matter. In a lecture on the challenges facing London as a leading dispute resolution centre, Sir Vos spoke about the role AI might play in helping resolve disputes more efficiently and suggested that AI could (albeit with some more learning) start to take minor judicial decisions, particularly in the commercial field. His statement was qualified; there would have to be controls including (a) that the parties understand which decisions are taken by judges and which by machines and (b) that there is always the option of an appeal to a human judge.

There are certainly some areas in which we may struggle to accept that a decision could be made by AI rather than a human. But no one knows quite what AI could be capable of and, given the rate at which it is learning (Sir Vos mentions in his lecture that  “when GPT-3.5 took the Bar exams not long ago, it came in the bottom 10%, but when GPT-4 took them just recently, it came in the top 10%”), it is perhaps not so hard to believe that in the not too distant future Sir Vos’ vision will be a reality.