Artificial lit, and crit

Artificial lit, and crit

An artificial intelligence is helping a novelist finish his sentences. Where’s the element of surprise?

artificial intelligence, Google AI, Goole, Syntax, computercalypse, express editorial
The AI suggests auto-complete fragments for sentences, a feature that Gmail users are quite familiar with.

Before a writer in Berkeley, California, started writing his second novel, he wrote an artificial intelligence (AI) to help him complete it. And after the novel is done, will it be reviewed by another AI, which will publish directly to the internet, without the mediation of editors? Or will the editors be AIs, too? Is the computercalypse upon us, then? Despite the excitement in the air, sadly, it’s old hat stretched to ten-gallon size. The AI suggests auto-complete fragments for sentences, a feature that Gmail users are quite familiar with. Write part of a sentence, the Google AI pops up a suggestion for the rest, based on the syntax of millions of users, and you only need to hit the tab key to accept it.

AIs, machine learning and deep learning are already in our midst. They map our route from point A to point B (Google), ensure inventory control and timely delivery of goods and services (Amazon), make games more real (Call of Duty) and steer us towards food, music and movies that they know we will like — even before we know it ourselves. In 2010, they caused the trillion-dollar Flash Crash, when machines allegedly run by a trader in his bedroom caused unprecedented volatility in futures markets. And we are on the brink of a new arms race of autonomous agents who will serve as proxies on the battlefield.

Did we say “who” rather than “which”? There lies the rub. In his last book, which is just out, Stephen Hawking has warned yet again of the risks of humans losing control over the development of artificial intelligence. If machines learn how to improve themselves and develop personality — and learning is AI’s core competency — they could put pedal to metal on a process of evolution that rapidly outstrips the biological world. It could end in a singularity, the point at which it surpasses human understanding. If it is reached, the power relationship between humans and their machines would be reversed, and we would be evolutionary history. The singularity lies in the unforeseeable future — somewhere in the latter half of this century — but AI is embedded in our lives already. Today, it is completing the sentences of a novel. Tomorrow, it could write a whole novel. It won’t be The Master and Margarita, but an AI could easily do a pulp Western or a romance. But could it do humour? That would be the Turing test for machines that write.