A Bored Editor & a Big Trash Can - Why Publishers will thrive in the age of AI

Currently, there is a massive blindspot in how we think about AI and future of media and publishing. The entire world is obsessed with the idea of text generation. People watch a chatbot spit out a 1,000 word essay in 4 seconds and think that the human publishers are on their last legs.

I mean if publishing were a contest to see who comes up with the most words or the cleanest sentences, then sure but otherwise, I feel the exact opposite. As sloppy, robotic content floods the web, I think demand for talented publishers is going to skyrocket. And that won't be because they can put the best words on a page but because they have exceptional taste!

The value of trash.

When you follow a publication, of course it could be because you love their writing style, opinions and wordplay, but a lot of times, it's just their taste. Even the sharpest prose and smartest opinions can't save a publication that picks uninteresting topics every day.You don't stick with a publication for what they put on the page but for what they throw in the trash.

Because everyone is so obsessed with how fast LLMs can generate text, we are overlooking the hardest part of running a publication - Curation! It is not writing the article that's difficult, it is looking at 100 different possible things to talk about today and having the taste to say no to 97 of them.

When supply is infinite, raw generation is worthless. Anyone with a prompt can fill a webpage (or tens of them) in less than a minute. What grabs your attention is the filter. You show up and stay with a publication because a human editor already went through the exhausting work of deciding what doesn't deserve your time.

Taste is very hard to replicate for AI

I think fundamentally, taste is very difficult for AI to replicate. Two humans can have similar tastes but never the same. LLMs are built to generate the most probable response by calculating the most expected next word. That's why their taste is super basic. LLMs rely on historical probability so they are only aware of what worked yesterday. They fail to bet on something brand new that they have no data about. A tasteful human publisher obviously filters out the noise but they also take risks on weird, niche or uncomfortable stories before the rest of the internet realizes they even matter.

It is good to be bored

Boredom is a very underrated trait that humans have. That fatigue is what naturally drives us towards new things. AIs don't get bored, they cannot feel the subtle cultural fatigue when a topic on the internet has gone stale by Thursday afternoon. While an LLM will gladly spit out 50th iteration of a dead trend because its math says people are still clicking on it, it takes a restless, bored human to look at what the world is talking about, yawn and find something better.

In older days, we needed publishers to tackle the hard problem of the time - writing and distributing stories. Today, we need them to ignore the slop and distribute stories that matter. Knowing what goes on the front page is important and I am really grateful that we have talented publishers to get the job done!