5 Comments

Software writers have been living with an AI helper called co-pilot for the past year or so https://github.com/features/copilot/

This feels pretty similar to the augmented version of AI for journalism: Nothing this thing produces could replace a human software writer and many of it’s suggestions are dumb. But it can help answer questions and occasionally makes suggestions that lets you skip a bunch of boilerplate.

Expand full comment

This coming Generative Content for SEO battle is interesting to me because increasingly I think that AI and search engines are part of a broader cyclical trend where information is organized, decayed by entropy, and then organized in a new form.

Search engines like Google and community efforts like Wikipedia were incredibly useful at their apex because of how cleanly they organized information. But then as those places became central, financially important institutions, they've been corroded by all sorts of forces and gradually taken to rot (I think Wikipedia is still pretty good, but it doesn't have that Alexandria feeling anymore).

ChatGPT is impressive, but it took a ton of bespoke labor to build all those datasets (in particular the reinforcement learning component). I wonder what happens when the paint starts to chip and all that training data gets dirtied by the fingers of commerce?

Expand full comment

Well said, Max.

I hope that BuzzFeed keeps the AI on the junk-food BuzzFeed site. BuzzFeed News kind of functions like the articles in Playboy. BuzzFeed is the nudity and centerfolds, but throughout its history Playboy also filled its magazine with longform literary journalism and muckraking. Its News site has really good longform stories and issues journalism.

I don't know if AI is ready for that kind of gathering and synthesizing data or quoting people.

Expand full comment

😌

Expand full comment
Feb 11, 2023·edited Feb 11, 2023

You're one of the first serious writers I've seen who hasn't been completely suckered in by the hype surrounding AI, and has voiced real skepticism about its potential.

What I'm reminded of, more than anything else, is the reality TV craze of the 2000s. It's hard to believe now, but there was once a time when it was seriously believed, by serious TV journalists and executives, that reality shows could largely replace scripted series. They'd no longer have to pay writers, and what's more, they'd be delivering audiences "real" drama and emotion, not something cooked up in a writers' room. The Fox network was leading the way into the future, and networks that relied on scripted series were dinosaurs. There were even experiments with reality movies like The Real Cancun.

Then came the 2007-08 WGA strike, and that theory was put to the test as the networks were forced to put all their scripted series on hold and run a slate of reality shows during the back half of the season. It was a disaster. There'd already been a growing cynicism towards reality TV by then, but after that cursed season of godawful midseason replacements and collapsing ratings, it hit a boiling point. After that, reality TV went from television's future to its open sewer, a wasteland of shows that were probably rigged and definitely trashy as hell.

My interactions with AI give me the same impression. It can do a passable imitation of a human writer or artist, but what it can't do is tell right from wrong. Not just in terms of morals, but in terms of discerning basic facts from falsehoods, though I feel that the two are connected. AI art has an "uncanny valley" feel to it where there are always certain details that it just can't pull off no matter what kind of art style it's been instructed to imitate, and like you said, AI articles and prompt responses are plagued with glaring factual errors.

AI won't be a complete bust. After all, reality TV did kill most of the tabloid talk shows of the '90s, and I suspect AI will do the same for SEO spam. But when it comes to journalism, I suspect that any outlet that replaces their writers with AI and wants to stay in business is gonna wind up hiring just as many editors to correct all the mistakes that their AI writer makes.

Expand full comment