[Kibae]

I really hope ChatGPT doesn't go with the advertising revenue model. I think the paid model they announced ($20/month) is a good starting point, but there's going to be significant drop-off from the mass market it serves today if there is no free tier.

I'm hoping they're able to figure out a model where the paying customers can offset the costs for the free tier for simple queries, and that the paid model is able to do more complex computations (e.g. Input a 300-page book and condense it into a 2-page summary; or integrate a search engine in applications like Obsidian).

[themodelplumber]

I like your ideas. I also think it'd be interesting if advertising models had a certain front-end algorithmic transparency which could be dialed in at a user's preference.

Example:

- I want the paid service, with no ads.

- I want the free service, with ads that use sampling model X based on my input.

- I want the free service, with ads that use metadata model Y based on my connection fingerprint.

- I want the free service, with ads that use another model, Z, based on source-convergence or some such thing.

- I want a discounted service, with ads of one of those types shown at period N of my choosing and based on the budget I specify.

It might also be interesting if providers like Google would let ad providers bid on your attention, by offering different experiences, algorithms, layouts, types of ad exposure, etc. with Google being an ad-experience aggregator.

In this way the door could be potentially opened to A) making ads more effective and B) helping users and e.g. parents of users set more comfortable boundaries with ad sources. You know what you're getting and you can change it to something else at any point.