Hacker News Reader

Comment by pjkundert | Hacker News Reader

It will be, if Elon implements effective agent-based clustering, such as with K-Means Clustering.

"Solving" content moderation is a losing battle -- there is no way to humanly moderate anything at the rate it can be produced, and Elon knows this. It's insane, so he probably won't do it.

But, allowing content to be produced, but making sure no real person ever sees it, until the content producer earns their way into "your" group -- now that can be automated. Also, then the police and FBI could use all the evil content to "do their job" (but, I'm not holding my breath).

Elon has access to Exascale hardware, so K-means Clusters with arbitrarily large dimensionality is at his disposal. The algorithms are linear in complexity, so should be parallelizable.

I figure he might have thought this through...

View on HN

I'd be interested to learn how you think kmeans helps either diversity or hate speech detection.

I don't think it helps with any of either, as people's behavior, the context, the language used and sensibilities change across cultures and over time. What you call hate speech now may not be in the future and may not have been in the past.

The best you could hope for is creating a fuzzy representation of the most visible problematic behavior and try to outrun model-world dealignment by constantly updating it.

The clustering responds to which tweets are up- and down-voted by the agent.

And, the agent is moved closer to the clusters they like tweets from, and further from the clusters they dislike tweets from.

The more diverse your likes (likes tweets in many other clusters), the broader and less restrictive your cluster weighting is -- the more variety you see. The more restrictive your selection, the less variety.

You decide how much of an "echo chamber" you're in. But, if you don't like $BAD_THING, and you downvote enough tweets of $BAD_THING, the less you'll see tweets by people in groups where they like $BAD_THING.

Recommender systems for news and social media have been tried, and I don't doubt that there may be niches where they succeed.

But large platforms have two massive problems:

(1) what people want is popular content, sometimes even content they would downvote. This destroys clustering by creating fuzzy centralized bridges and erodes the usefulness of recommenders

(2) people over time have lost trust and interest in highly personalized feeds, see facebook or the revolt against instagram's feed sorting.

Two cases that seem to work for now are music recommendation and tiktok. I'm not holding my breath though, because spotify might end up driving people away with too many monetized podcasts and tiktok could succumb to the generational migration. But we'll see!

Sentiment analysis can solve the first. If the algo notices that someone only engages negatively with another cluster, cut it off; they're not entitled to pollute that.

And the second point was a mostly abstract intellectual debate in the early 2010s, but people have proven that they absolutely prefer to stick with their own, and have close-to-zero tolerance for dissent or disagreement (see, "the hivemind"). Twitter is far more prone to this than Facebook, since it's founded on communities (i.e. clusters) rather than just friends. Twitter just needs to stop putting junk into people's feeds; and silo them better to reduce harassment.

That would also reduce polarization, since a yuge cause for polarization (far-right, antifa) is a knee-jerk reaction to the very worst content from the other side. Stop promoting that, and you'll stop the "Brainwashing Of My Dad" effect (the film), where a conservative from Texas gets upset because of bathroom laws in California, or where a liberal gets upset because of a few 4chan trolls.

It's the psychological phenomenon of "enmeshment": remove the boundaries between people, and you make their relationship very toxic and conflictual. Enforce better boundaries, and their relationship will be far healthier. Twitter promoted the former; Musk can avoid most of the upcoming "hell" being predicted by doing the latter.

I think polarization is likely to be much more extreme if you silo people off into like minded groups.

People will get bored, eventually, slumbering in a Pablum of like-mindedness.

How does this work with things like "ratio-ing" or "dunking" or libsoftiktok that intentionally take clusters of media from one group and shoving it into their group explicitly to hate them?

I'm not sure that shaking your head and laughing at someone's crazy-talk is "hating" them... Perhaps that's part of the problem.

If you don't want people chuckling at you, don't talk crazy.

If you really, really get annoyed at something and mash the downvote button, it'll eventually go away, and you'll be left in your warm, cozy bubble.

Everyone wins!

Isn't this just shadow-banning, which is already a thing people believe is the same as censorship and enumerate numerous conspiracies about?

If you choose to downvote lots of $SOMETHING, you'll move further away from groups where they like $SOMETHING, and see less tweets about it probably.

Your delicate sensitivites about $SOMETHING don't affect anyone else.

If anything is clear from the texts that came up during discovery, it's that Elon Musk hasn't thought anything through with this deal.

Long Covid brain fog? Marijuana-induced schizophrenia? Has he just been stupid and lucky all along? I haven't spent enough time thinking about this to figure out why.

How does that help? The whole problem is that twitter provides a platform to legitimize hate speech by spreading it to like-minded people and thereby emboldening them. clustering does exactly that, but shields bad actors with a cloak of secrecy.

There's no secrecy. The FBI can (and should) infiltrate these groups (just like their tweets, and you're in), and then arrest them for their illegal behavior. And, limited in-real-life moderation of illegal posts can be focused on where the crazy people hang out. It should be a win.

Advertisers don't want placement next to "bad" content, whether or not the in-group likes that content.

(That's probably a big reason why he wants subscription fees to take over advertiser revenue dominance.)

Of course. A wise advertiser would select a subset of the K-Means Clusters to advertise to. Like, probably the "not insane" ones.

oh no, not k-means clustering

My life was going fine until my tweet got too close to Agent K's centroid.

I think it'll work more like this.

There are people who enjoy reasoned debate, even if they don't agree with the counterparty. They won't downvote those tweets. Their cluster weight will be gently moving in the direction of others like them, and its weight will let through a wide variety of other gently weighted groups.

Others are compelled to screech and foam at the mouth if someone argues for a $WRONGTHINK - even in jest or as a devil's advocate, and will downvote them instantly. Their weightings will be heavy, and will exclude most other groups. They will see only what they want to see, and won't see what they don't want to see -- reasoned debate from groups they disagree with, for example.

This is a personal choice, and mirrors physical group dynamics.

If someone "in your group" starts talking crazy, they will begin to separate from the group (by mutual down-voting), and migrate to a different group. You (and the rest of your group) won't be affected much.

Reply on news.ycombinator.com