I feel like every federated platform will inevitably add some form of machine learning or algorithmic discovery if they want to grow beyond 1M MAU
I don't think it's a bad thing if done properly
@dansup as long as the recommendation algorithms can be selected and disabled by the user, I don't have a problem with it.
@dansup algorithmic discovery is such a huge category though, and completely different from machine learning
@dansup They might even have multiple algorithms you can choose from, or allow api access to provide your own algorithm, or join up with some sort of federated sites algorithm, or whatever. The key differences with current platforms should be 1) they will offer users options with transparency, and 2) algorithms will be to the user's benefit and not be built around the goal of offering users up to advertisers.
@dansup I agree that a single federated instance would likely need to empower users with some type of advanced feed controls in order to grow beyond 1M MAU.
And I would welcome such a feature, even on my own instance.
But I actually wonder why an instance would want to grow beyond this threshold. The profit motive isn't really there. Adding feed controls and "popular content" would make it more welcoming for new users. But it would cost more money and those users usually don't bring money.
@dansup instead of server-side algorithms or ML models, I'd say have an option for users to create and modify weight models for content - toots by this user gains 1 point, this hashtag gains 2 points, etc.
Which can then be adjusted per interaction by the software, but would be still open to the user for their inputs. I'd personally love that.