Marcus du Sautoy: How AI is learning to write, paint and think

A harmonious sigh of relief could be heard across the Watershed auditorium on Friday lunchtime as a sprightly Marcus du Sautoy began his talk by reassuring listeners that he was there to promote the exciting potential of AI instead of contributing to the dystopian discourse that surrounds the future of technology. Whether or not the audience left feeling comforted is another question.

The Professor of Mathematics began his lecture by outlining the fundamental shift that has taken place in the world of coding in the past ten years. Previously, Artificial Intelligence had been coded with set responses to particular stimuli. The best example of this is the computer Deep Blue’s victory over the world champion, Garry Kasparov, during a game of chess at the end of the 90s. Now, however, code is being written in a bottom up manner to mimic that of the human learning process. This means that AI learns from its own mistakes; it updates and grows continually stronger. As such, AI has the ability to be independently creative - a notion that is understandably alarming to the previously untouchable arts sector.

Du Sautoy carefully selected examples from across the arts to demonstrate how AI has surpassed human creativity. Audiences were asked to vote on which painting they thought was the original Rembrandt and which was generated by AI or select which poem out of three was written by a human. For each scenario opinions were split almost fifty-fifty effectively demonstrating how far AI has come.

It is interesting that in almost every example of AI’s success the identities of the creators are concealed in order to demonstrate the homogeneity of man and bot. This is presented as a great feat of technology since the ultimate objective is to align the two as closely as possible. However, scientists aren’t imitating a perfect model. Humans are prone to prejudices, inconsistencies and errors. Whose responsibility is it to code morals into AI? Marcus du Sautoy recounted a time when he had witnessed a black, female scientist unveiling robots. The machines failed to acknowledge her until she placed a white mask over her face. They had been programmed to respond only to white males highlighting a critical moral issue that mustn’t be overlooked amongst the fanfare of success.

This led me to consider the role of the humanities in a future world of AI seeing as artistic output could be reduced to a code. Through the help of the humanities one is able to see beyond the art: to reflect on the process of creation, question how it connects to the surrounding world and inject the human player back into the narrative of its conception. I think the answer lies in the very root of the word. In a future of growing uncertainty, fuelled by the rise of the machines, we mustn’t lose sight of what it means to be human.