I began my career in artificial intelligence more than two decades ago. Back then, explaining what I did at dinner parties would earn me a polite smile and a couple of jokes about R2-D2 before the conversation moved on. Not these days.

Now, as soon as those two letters – AI – leave my mouth, a conversation is in full swing, and everyone has an opinion, ranging from anxiety about job losses to “ChatGPT makes me so much more productive”. I am often caught in the middle, trying to bring a sense of balance.

It was curious, then, that when I was named in these pages a few weekends ago, that sense of balance was lost. In an opinion piece for this masthead, comments I had made during a talk on one of AI’s many flashpoints – AI and the arts – were used as a jumping-off point. I came across as wholly enthusiastic about AI, which is not the full story.

It’s true that I used many “unbelievable” examples of how AI is transforming the arts. There was the example of a team that used generative AI to create a reasonable Taylor Swift video in less than a day. And the UK TV show that superimposes video of celebrity faces on the bodies of actors to put those actors – presumably without their consent – in situations they would never normally find themselves in, all in the name of comedy.

And the arts is no lonely outlier here. If, like me, you are still on a high from the performance of the Matildas, you might be interested to know that AI is everywhere in sport. And it’s not just for those obvious things such as checking if a ball has hit a line or strayed out of bounds. There is now an AI that can automatically add sound effects to the bounce of a tennis ball, making it possible for vision-impaired people to follow a match.

Loading

I am what I’d call an “AI positivist”, and I am genuinely excited by these examples. They are things we could have only imagined when I started out in AI.

But AI positivism isn’t AI evangelism. There are plenty of those out there, and I’m not one. Rather, I believe AI, if implemented responsibly, can bring significant new benefits to society.

My definition of AI positivism, however, relies crucially on the phrase “if implemented responsibly”. And the “if” is no small caveat. The current AI zeitgeist is over-hyped, and there are very real dangers to AI that we need to address. They include copyright infringement in AI-generated art. Then there is AI-generated misinformation – even “hallucinations”, where sentences assembled in a statistically probable way nevertheless introduce details that are not true. And there are concerns about whether athletes, for instance, have enough control over their own performance data.