communitynotes

> When you use the term "open-source" in this context, what do you mean? The code for its implementation is literally available on GitHub [0] under the Apache License 2.0. [0]: https://github.com/twitter/communitynotes

I know there has been a lot of ink spilled trying to persuade that Technology can't solve our deeper problems and Technologists are too optimistic about having real-world impact etc. etc. But I think community notes (the model, not necessarily the specific implementation of one company or another) is one of those algorithms that truly solve a messy messy sticky problem. Fact-checking and other "Big J Journalist" attempts to suppress "misinformation" is a very broken model. 1) It leads to less and less trust in the fact checkers which drives more people to fringe outlets etc. 2) They also are in fact quite biased (as are all humans, but it's less important if your electrician has socialist/MAGA/Libertarian biases) 3) The absolute firehose of online content means fact checkers/media etc. can't actually fact check everything and end up fact

The X Community Notes feature is open source. That is at least a step in the right direction, even if the feed algorithm is still closed. https://github.com/twitter/communitynotes For better or worse, X / Twitter was never really about communities. It has always been focused on allowing individuals to broadcast their thoughts. If you want a moderated community discussion then there are much better platforms for that. We do have neutral mediums for free speech: HTTP and SMTP.

It is open source: https://github.com/twitter/communitynotes

X | Machine Learning Engineer - Community Notes | SF, SJ, Seattle, LA, NYC | All Levels The role: build scalable production machine learning systems and infrastructure, and improve our open source algorithm: https://github.com/twitter/communitynotes Community Notes has been discussed here on HN a number of times e.g. here: https://news.ycombinator.com/item?id=37292041 Apply here: https://twitter.wd5.myworkdayjobs.com/X/job/San-Francisco-CA... There are many other positions at the company too, see: https://careers.x.com/en

> Good point that the algorithm tries to compensate for the perspectives, but I’m sure it still comes down to a popularity contest. It isn't clear if polarization score have only one dimension where it would be great to capture US culture wars but fail to capture nuances outside of that or if it's more complex than that. > Generally, the platform determines the algorithm of which note wins, so that’s centralized. The algorithm depends on what kind and how many users vote and how. Those users exist on the platform which requires registration and can deny any given user. Centralized. Yes, not perfect but still better than the traditional media oligopoly. > Further, no guarantee that the actual algorithm in production matches the one made public, but I guess they have no reason to lie here. The algorithm and the data are open. It is reproduc

The data are publicly available so I suppose anyone can do this https://github.com/twitter/communitynotes

it's neither of those which leads to your own lack of knowledge of the system. you can see the source code of the algorithm here: https://github.com/twitter/communitynotes there's a number of factors that goes into the rankings, like taking into account which other community note users rated it helpful and how they rated previous notes. also simple usage of community notes would have revealed it was at least more sophisticated than the prison of two ideas you presented.

>Did they publish the details of these systems? I didn't look into the details myself, but there is a 24-page paper linked from this page: >Details of our methodology and past findings can be found in our paper. https://communitynotes.twitter.com/guide/en/under-the-hood/g... https://github.com/twitter/communitynotes/blob/main/birdwatc...

Paper explaining it in detail: https://github.com/twitter/communitynotes/blob/main/birdwatc... It's somewhat adversarial, with the idea that if people who have disagreed in the past agree on a note, then it's probably true.

> Honestly , the product has become much better since he took over. Indeed, having internet prostitutes with a blue tick as top comment on almost every tweet, what's not to like… (at least I'd rather have Onlyfan sex-workers on top than Musk's fan base but still). > The community notes were developed before the take over but Twitter apparently opted for centralised moderation and did not introduce the feature. Community notes started in 2021, as a pilot called birdwatch who was basically the same thing and was already open source and decentralized[1] and it was slowly being expanded in scope, the fact that the expansion culminated roughly at the same time as Musk taking over is essentially a coincidence. [1]: https://github.com/twitter/communitynotes/commit/9441a8205f4...

The code is open source: https://github.com/twitter/communitynotes The notes and rating data is released to the public every day: https://twitter.com/i/communitynotes/download-data Feel free to run the matrix factorization code on the data, and then try to interpret the resulting latent dimension it finds for yourself! And also, you can read the code to verify that it really is running a matrix factorization rather than hardcoding a particular left/right split.