Sloppy thinking

AI-enabled copy is much better than many think

“No slop in our shop” was reportedly the message from Business Insider reporters to Mathias Döfner, the ceo and co-owner of their German parent company Axel Springer, when he visited their offices early this week. He has been a great proponent of the use of AI in his newsrooms; his reporters, at least at Business Insider, aren’t as keen, it seems.

The long-term effect of AI on journalism is something we’ll find out in a few years’, or even decades’, time, but writing off its output as “slop” is very much the insult of the moment. In my view, this is lazy, hypocritical and frankly naive.

First, lazy because users of the term are putting everything produced by AI into the same bucket. I’ve seen “slop” used to describe everything from Donald Trump’s shit bombing post to video of horrified mothers rescuing their bear-feeding children to standard AI-drafted articles. They are certainly not the same thing.

Second, hypocritical because journalism needs to get its own house in order first. We are certainly producing a good amount of slop ourselves without a hint of intelligence.

Take The Times’s retraction of a juicy interview that it thought it was conducting with Bill de Blasio, the former mayor of New York, when actually it was speaking to Bill DeBlasio, a chatty wine importer from Long Island. “Humiliating” was the verdict of the paper’s head of standards.

The episode was seized upon by American media commentators who couldn’t imagine how such a mistake could happen. Sadly, anyone with any experience of Fleet Street knew exactly how and why it happened. There’s still too much of this slop about.

Third, naive because everyone who has been working closely with AI over the past six months will have noticed quite how much better the models’ output is getting. The rough edges, stylistic tics and hallucinations seem to be reducing at the same time as their capabilities are increasing.

An example is the interview conducted by ChatGPT of David Caswell, the researcher and consultant on AI in news, which we published in Tomorrow's Publisher this week.

It was about David’s undertaking of a journalism futures project using AI agents which aimed to replicate one conducted by humans the year before. It’s fascinating reading just on that account but also it’s notable that the interviewer does a stellar job based on a very simple prompt. David said it was “the most penetrating and perceptive interview I’ve ever done”.

As ever, we must remember that these are very early days indeed in this particular revolution.

So whenever I hear people in the media railing against AI slop I know not to listen to them.