Yesterday Google's artificial intelligence (AI) development company, DeepMind, announced another major step forwards for automated machine-learning technology, having produced a programme able to learn how to play – and win – the Chinese game of Go, without human input. The original AlphaGo programme was programmed to learn how to play Go by analysing moves from thousands of games played by humans. But DeepMind's latest AlphaGo Zero programme was given a blank Go board and just a set of rules. It then learned how to play by playing itself. Within just three days it had mastered the game to an astounding degree, having beaten the original AlphaGo programme by 100 games to zero.
AlphaGo – not just a game
There has been a huge amount of hype surrounding AI, robotics and automation, most of which is undeserved and exaggerated, at least according to this BBC news article, which acknowledges, however, that AlphaGo Zero “is at the other end of the spectrum – proper peer-reviewed science with a real advance in computer intelligence.” DeepMind's chief executive, Demis Hassabis, said AlphaGo Zero could eventually have other applications – but cautioned that AI development is still in its early stages.
Automation of tasks, not jobs
It's nonetheless a topic that has captured the attention of financial professionals and has been high on the programmes of this Autumn's financial conferences. EuroFinance's International Treasury & Cash Management event in Barcelona earlier this month had a number of speakers on the topic, including scientific adviser on AI, Dr. Adam Rutherford, who reassured the audience that humans will always be needed to work with intelligent machines and ensure they are carrying out analysis specifically tailored to human needs. And at the AFP's annual conference this week, Julia Kirby, senior editor at the Harvard University Press, noted that it will be tasks that are automated, not jobs, providing further reassurance for financial professionals.
90% of treasury jobs could go to AI
However, Jean Furter, treasurer for Brocade Communications, told the AFP's Andrew Deichler that he believes 90 per cent of the tasks carried out by treasury professionals will eventually become automated, which could lead to a lot of consolidation in the financial function.
Furter could be nearer the mark, particularly as AlphaGo Zero underlines the capabilities AI now has for independent learning and designing software that is better than software designed by humans. This article, by Elena Mesropyan published in Let's Talk Payments, underlines how Airbnb has been using automated machine learning to develop sophisticated algorithms and systems used for benchmarking data, for diagnostics and for exploratory data analysis, pre-processing of data and for model selection. As we are now reaching a stage in which systems can produce other, better, more efficient systems and software, this could have several implications. For a start, software developers might feel even more under threat than those working in finance and treasury. And companies might have to reassess their employees' skillsets in some areas, with a greater focus on recruiting and training machine-learning experts. As has been often repeated during this season's treasury events, financial professionals are strongly advised to develop a broad skillset, as the more specialist financial roles are those that are most likely to be automated.
Like this item? Get our Weekly Update newsletter. Subscribe today