Artificial intelligence

It feels, pleasingly, like a scene from a cerebral James Bond film, or perhaps an episode of Slow Horses. I am in a shadowy corner of a plush, buzzy Soho members’ bar. A mild December twilight is falling over London. Across the table from me sits an old acquaintance, a senior English barrister, greying, quietly handsome, in his mid fifties. And he wants to speak anonymously, because what he is about to say will earn the loathing of his entire profession. Let’s call him James. I’ve known him for a few years, and over these years we’ve discussed all kinds of things, from politics to architecture to the misfortunes of Chelsea FC. We’ve also discussed technology and AI. James’s views of AI were always like his politics: centrist, clever, moderate, sceptical.

In search of coffee on my way to work the other day, I stopped short mid-way into a branch of a popular coffee shop when I noticed the digital ordering screens. Nothing will lose my business faster than being asked to queue twice and do the work of someone else for something simple. But these ordering screens seem to be becoming ubiquitous in our towns and cities, forcing those of us who have actually come into the office, likely to sit in front of a screen, to spend our lunchtimes also staring at a screen scrolling through options, when there is an actual human being standing behind a counter a few feet away with nothing to do.

There are several class actions going on against developers of Large Language Models. Jodi Picoult, George R.R. Martin, John Grisham and several other well-known authors are among those engaging in long-drawn-out lawsuits with tech companies such as Meta (who developed the chatbot LLaMA), OpenAI (who developed ChatGPT) and Google DeepMind (who developed Gemini). These companies, without seeking permission (imagine!), used books, newspapers, websites and other text sources to generate datasets to train their machines. The lawyers for these tech companies claim it was ‘fair use’. No one actually copied and resold anything, they say; it was used to train, and only to train.  One of my novels was ‘scraped’ by Meta, and features, in a tiny way, in the lawsuit.

Early in 2020, inquiries for our speech-writing services were arriving in their droves. From Westminster to Washington, weddings to wine tastings, people needed our help. We cancelled our weekends and prepared for life without a mortgage. Covid gave us our weekends back. And all the other days. Yet when parties and events returned, a significant chunk of our clients did not. It was weird. But this wasn’t a vaccine complication, just a new player in the market. Previously we’d only had to win business against other humans. Suddenly, we were faced with a competitor able to provide speeches for any occasion in seconds. ChatGPT was doing to us what PornHub had done to the top shelf in the local newsagent. We weren’t the only ones.