4 ms·
The problems with Elon Musk’s plan to open-source the Twitter algorithm
- whywhywhywhy 5y agoTelling us “you wouldn’t understand it” increasingly feels like an excuse. Here’s a wild idea, maybe put some effort into building a window into the system that does make it understandable. Have a strong feeling you at least have made tooling like that for your internal teams to understand what it’s doing.
- noduerme 5y agoMastodon explicitly doesn't follow these kinds of opaque algos for recommending tweets. The risk that someone will create a "knock-off" twitter? Give me a break. Anyone could do that at any time, there's nothing special about Twitter except the enormous network effect. The only risk I can see in open-sourcing the code is that it would make it easier for people to game the system and attack it in the short run. But opening it to scrutiny and review will make it more robust and less susceptible to attack in the medium-to-long term.
- usrn 5y agoIMO: Mastodon is already a great knock-off Twitter. There's much more conversations between peers compared to Twitter which feels like celebrity RSS with a comment section.
- nonrandomstring 5y agoWhat a badly written and dishonest article. > Allowing Twitter’s algorithm to be visible to others, and adaptable by competitors, theoretically means someone could just copy Twitter's source code and release a re-branded version. Large parts of the internet run on open-source software most famously OpenSSL a security toolkit used by large parts of the web, which in 2014 suffered a major security breach. Sounds like the desperate, misinformed anti-open rhetoric from the mid 1990s. The real issue with Elon opening the algorithms, which I commented on a few days ago [1], is skilfully avoided. Those algorithms, and their version histories, encode many dirty secrets that will enable an astute analyst to walk back through the timeline exposing prior politically influence. The actual source code is irrelevant (for Kerckhoff-Shannon reasons). It's the parametric data associated with it that tells a thousand tales. [1] https://news.ycombinator.com/item?id=31183198 https://news.ycombinator.com/item?id=31183198
- karmakaze 5y agoI almost stopped reading there, as it didn't seem honest or accurate. It does get somewhat better: > For Cobbe, the risks outweigh the potential benefits. The computer code doesn’t give us any insight into how algorithms were trained or tested, what factors or considerations went into them, or what sorts of things were prioritized in the process, so open-sourcing it may not make a meaningful difference to transparency at Twitter. Meanwhile, it could introduce some significant security risks. So really this is saying that opensourcing a selection of algorithms isn't enough. Whenever I see something that says this is a bad idea or can't be done, I see the reasons given as a TODO list of how to actually make it work. What we want is transparency of changes. What's the goal, what's being measured, what's the impact of a change.