My first request: never show my video or pictures of violence. It’s unbelievable how much we’ve normalized “here’s video of an actual murder where a real human being dies”.
How? If the algo is truly open source and each person can prompt Grok however they like in order for Grok to tweak that open-source algo to manage their individual X feed then that seems like a clear increase in transparency.
We are focused on generated feeds based upon structured queries and semantic search.
It's a huge construction site, but it's already pretty cool and will soon be 🔥.
Forgot to mention that we are also planning on adding an API, so that you can use your own, as well, but figuring that out is more difficult, as we won't know what information the AI is sending us and we don't want to turn the results into a a bunch of chatbot drivel.
For the open-sourcing I think he's talking about X's recommendation algorithm for the feed, which is not Grok. You tell Grok what you want and Grok customises that recommendation algorithm based on what you told it. That algorithm is something a human can parse.
Grok itself is not being open sourced, and even it were human brains can't get their heads around these models anyway.