Publishers need to train their own LLMs for their readers

I’m assuming you’ve used ChatGPT. If so, you’ve probably run into its very annoying super ego. For example, you’re doing some research on nuclear power and it hectors you with grade school morality about the dangers of this and that.

It’s quite annoying.

But there’s a value in that, which is that it’s possible to layer a point of view on top of AI responses.

That’s actually a good thing.

As I’ve mentioned a few times before, tools like ChatGPT will undermine the current search engine model, where I ask a question and Google gives me some homework to do.

“Here, read all these articles and you might find what you’re looking for.”

That was very useful in its time, but now it’s rather quaint. When I ask a question, I want an answer.

The next crucial step is that I might want an answer from a particular point of view.

Let’s say I’m a Lutheran seminary student. I might want to have an answer that’s consistent with Lutheran doctrine. I might even want to get more precise and ask for Philip Melanchthon’s opinion on a particular idea. (Philip was a disciple of Luther.)

You can do that with ChatGPT. I just tried it, in fact, and it did a remarkably good job. It also bypassed that annoying super ego problem. It didn’t keep reminding me that there are other views and blah blah blah. I wasn’t asking about other views. I was asking about Philip’s views, and that’s what it gave me.

Here’s the point for publishers. Let’s say you run an investing newsletter. There are all sorts of different strategies and mindsets and approaches to investing. If you hook up AI to your site, you want it fine-tuned to represent your brand with your voice and your approach.

That’s what publishers need to start doing. Not only do you need to feed your content into your own instance of a large language model. You also need to tinker with the super ego of the AI so that it learns to respond in your voice, with your perspective.