Technology

It already does, though not in the individualized manner he's describing.

I don't think that's entirely a bad thing. Its current form, where priority one is keeping advertisers happy is a bad thing, but I'm going to guess everyone reading this has a machine learning algorithm of some sort keeping most of the spam out of their email.

BlueSky's labelers are a step toward the individualized approach. I like them; one of the first things I did there is filter out what one labeler flags as AI-generated images.

No thanks.

Empowering users or just handing them the keys to their own echo chambers? Innovative but fraught with potential downsides.

🐱🐱🐱

Yeah but my madlibs generator might get confused without a /s

At least 'AI' has thick skin?

I disagree, but I think he's close. The future of moderation should be customizable by users, but it needs to be based on human moderation. Let them pick their own moderators, and fine tune that moderation to their liking, and give them an option to review moderation and make adjustments.

I think that he's probably correct that this is, in significant part, going to be the future.

I don't think that human moderation is going to entirely vanish, but it's not cheap to pay a ton of humans to do what it would take. A lot of moderation is, well...fairly mechanical. Like, it's probably possible to detect, with reasonable accuracy, that you've got a flamewar on your hands, stuff like that. You'd want to do as much as you can in software.

Human moderators sleep, leave the keyboard, do things like that. Software doesn't.

Also, if you have cheap-enough text classification, you can do it on a per-user basis, so that instead of a global view of the world, different people see different content being filtered and recommended, which I think is what he's proposing:

Ohanian said at the conference that he thinks social media will "eventually get to a place where we get to choose our own algorithm."

Most social media relies on at least some level of recommendations.

This isn't even new for him. The original vision for Reddit, as I recall, was that the voting was going to be used to build a per-user profile to feed a recommendations engine. That never really happened. Instead, one wound up with subreddits (so self-selecting communities are part of it) and a global voting on stuff within that.

I mean, text classifiers aimed at filtering out spam have been around forever for email. It's not even terribly new technology. Some subreddits on Reddit had bots run by moderators that did do some level of automated moderation.