the-algorithm-ml VS benchmarks

I said “parts of the recommender system code.”

This is the kind of highly emotional reaction that’s not helpful.

Yes, I am quite familiar with building ML models, both training and building my own for which I’ve been paid large sums of money, and I’m here to tell you that you don’t know what you’re taking about.

There’s so much more information about an ML system than just the trained model that is important for understanding the effects of the system on a society, and its legal, ethical, and social ramifications.

Just seeing the type of RS being used, the ranking approach, and the information on SimClusters is enough for RAI folks to start to understand the ecosystem effects and how that can show up downstream in social effects.

https://blog.twitter.com/engineering/en_us/topics/open-sourc...

Blog summarising the change: https://blog.twitter.com/engineering/en_us/topics/open-source/2023/twitter-recommendation-algorithm

People who don't share your interests (or at least what Twitter thinks your interests are). This blog post explains it in detail.

Twitter's announcement | Main GitHub Repo | ML GitHub Repo | Engineering Blog Post

where? I searched through the repo and couldn't find it.

GitHub - kostya/benchmarks: Some benchmarks of different languages

In all the language comparisons I've found over the years, Python consistently comes out slightly slower, for example:

https://github.com/kostya/benchmarks

Bearing in mind these are probably not even using YJIT, which makes Ruby considerably faster in some scenarios.

Also, here is another benchmark: https://github.com/kostya/benchmarks