What is the cost?

Caswell: In the last decade it was very expensive. It was very difficult: You need the data, you had to build a data warehouse, have an enterprise deal with Amazon or Google cloud, you had to hire data scientists, to have a team of data engineers. it was a major investment. Only the BBC, the New York Times, this level of organisations could really afford it.

That’s not true with generative AI. You can run news workflow through interfaces that you pay 20 dollars a month. You don’t need to be a coder. All you need is motivation, enthusiasm and curiosity.

There’s lots of people in news organisations that would not have been involved in AI in the past because they did not have the technical background and now they can just use it. It’s a much more open form of AI: both smaller newsrooms can do a lot with, and more junior individuals in more established newsrooms can do a lot with. I think it’s a good thing, but it’s also a disruptive thing. Often the internal politics in newsrooms are disrupted by that.

At what stage of AI are we at?

Caswell: AI has been around since the 1950s. But AI for practical purposes appeared with ChatGPT. It’s going to be quite a while — years — before we really understand how to use them for valuable things. There are so many things that you can do with them.

The risk to journalism is that other organisations, start-ups, tech companies will do things in news faster than the news world itself. Lots of start ups have no editorial component at all. They are swiping the content of news organisations, some are covering niches: they are monitoring press releases, social media channels, PDF from reports.