This is the second part of a two-part blog where we look at the idea of the Centaur – how humans can add value to the large-scale computational processes such as Big Data management and processing, AI and machine learning, and the algorithms that come out of these, which allow for automated decision making in the Age of Information.

After I read the James Bridle essay discussed last week, I stumbled upon an excellent series of articles by Australian radiologist and medical AI researcher Luke Oakden-Rayner titled The End of Human Doctors, which poses the question “Will computers make doctors obsolete, and if so, how soon?” Anyone who’s interested in AI and machine learning should read them. In the third essay on Understanding Automation he explores this notion of augmentation of computing capabilities by people. He draws out the genealogy of this particular piece of this piece of wisdom, one that’s edging into the ‘received wisdom’ zone, with its source in Gary Kasparov himself:

There is one main piece of evidence that gets regularly cited:

Humans have lost to computers at chess for decades, but hybrid human/computer teams beat computers alone.

The idea is that humans and machines have different skillsets, and working together they are better than the sum of their parts. I believe this idea was first introduced by Gary Kasparov, and then popularised by Tyler Cowen in his book “Average Is Over“. It has since been repeated by many other authors including Erik Brynjolfsson and Andrew McAfee in “The Second Machine Age“. Both books have been highly influential in the discussion on automation, and this idea has been repeated over and over again in dozens of articles on automation in major news outlets.

The End of Human Doctors – Part 3: Understanding Automation by Luke Oakden-Rayner

But, he goes on to say, in correspondence chess, where you assume all games are computer+person competitions, the draw rate for some time has been approaching 100%. Assuming that all chess players haven’t suddenly become equal that can only mean that the automation has now reached a point where it effectively renders the benefit of the human element null. Hybrid or ‘advanced’ chess tournaments have died out. Too boring. Bye bye centaur, hello horse. Chess player takes up a career in professional tiddlywinks or goes to be melted down for glue at the nearest job centre.

The conclusion Oakden-Rayner makes is this:

At the very least I think the argument that “human/computer teams will always be relevant” is unsupported by the experience in chess. It seems more likely that there is a transition period where systems with identifiable weaknesses can be augmented by human assistance (this “brittleness” was much more of a problem before deep learning..).

The use of Automatic Speech Recognition for the purposes of same-language subtitling is in just such a “brittle” transitional period, with ASR becoming useful for some tasks, but unsuitable for total automation of the subtitling process in any sort of regulated market.

Nevertheless the last couple of years, with the impact of deep learning in the ASR area, has seen rapid improvements in accuracy, and it’s clear the application of ASR in subtitling is going to grow.

During this process the centaur image should not be deployed too easily. We must go beyond its central image  and think carefully about where people will have most value and how we can help them realise it. That may mean a period of augmentation – the time of the centaur – but it will also mean training and giving people the new skills to move outside those areas of automation, not just enable them to work within it. Doing this should ensure the imagination and creativity that came up with centaurs in the first place. Actual centaurs.

Tom Wootton, Product Manager Access Services