As we've already seen from twitter releasing the algorithm: it's useless & irrelevant without the weights.
For example, Twitter releasing the algorithm showed us that certain specific people were getting artificially boosted.
We also saw that the topic of the war in Ukraine was getting special treatment.
Yes, we all guessed that was happening, but without the code it was just supposition that Twitter could deny.
It did? Did I miss something here? I don't remember seeing that. There was certainly code that tallied certain specific people and topics, but I don't see an indication of what was done with that tally.
> To help assuage Musk’s concerns, Platformer reports that Twitter’s engineers created a way to “tweak” the site’s ranking system when they noticed a high-profile user’s engagement dropping, ensuring “that tweets from those accounts were always shown.” That’s part of the reason why users saw Musk all over their timelines in February, as it essentially exempts these accounts from Twitter’s algorithm (although Platformer says Musk’s boost was different from what all the other users receive).
https://gizmodo.com/twitter-algorithm-aoc-ben-shapiro-cattur...
> In February, Twitter rolled out a system to boost Musk’s tweets over everyone else’s and began boosting the tweets of these VIPS, albeit to a lesser extent.
https://gizmodo.com/twitter-musk-ukraine-crisis-open-source-...
> Misinformation is highly down-ranked
> Anything that is categorized as misinformation gets the rug pulled out from under it.
> Surprisingly, so are posts about Ukraine.
...
> Late on Monday, Gupta told Gizmodo that “upon further investigation” he and his fellows looking into the Twitter code found that it only applies to Twitter Spaces, the platform’s live meeting feature.
I though that perhaps you were talking about something new.
The Ukraine issue was only discovered due to the code disclosure.
Two, Twitter claims that whole area was for telemetry.
This comment does a massive disservice & deeply injures the cause for real transparency. It does it by making a very cheap shot at a very dumb situation. To say that this makes not having the models OK or interesting or useful? I highly protest. This is a sad distracting sideshow & letting oneself be lured & baited into thinking this little morsel is filling or informative is... Ughh I just can't no, that's just so low a position.
Yes, it will, because with weights you can run it against constructed data, and that kind of experiment with the actual algorithm (including weights) is a much better source of practical understanding than abstract analysis of the algorithm without weights.