[17:27:23] 10Machine-Learning-Team, 10Analytics-Radar, 10SRE: Kubeflow on stat machines - https://phabricator.wikimedia.org/T275551 (10razzi) [21:06:55] accraze The research team wanted to know if we wanted to talk about AI model governance, or at least the very start of it, to the Research Showcase [21:07:23] I said yes as long as they didn't mind mostly our initial thoughts and then invite the community to comment and give feedback [21:08:14] sure yeah that sounds great! [21:08:47] I also have it on the agenda for the ML sync, but that might not be for four weeks [21:09:12] Unrelated but I also just did an impromptu livestreamed ML office hours using Twitch [21:09:42] it works really well, with no notice, maybe 30 folks showed up [21:10:27] I am going to continue to do them on a regular schedule to give folks a chance to learn more about what we are working on and where they can plug in [21:10:35] chrisalbon: thats awesome, we should totally do more of that [21:11:35] There was alot of interest https://twitter.com/chrisalbon/status/1367536299876360194?s=20 [21:11:56] Also, YouTube is probably a better platform https://twitter.com/chrisalbon/status/1367540693774737408?s=20 [21:12:18] However youtube requires you wait 24 hours after registering an account to livestream, so I tried out twitch [22:36:36] 10Machine-Learning-Team: Experiment with on-wiki model documentation - https://phabricator.wikimedia.org/T276398 (10ACraze) The [[ https://arxiv.org/abs/1810.03993 | model cards paper ]] recommends having the following sections: - **Model Details** - Who built the model - Date developed - Model ve... [23:47:55] 10Machine-Learning-Team: Model Reporting - https://phabricator.wikimedia.org/T276397 (10ACraze) The historical ORES model reports are in version control for each repo (editquality, articlequality, etc.). We should be able to get model version, fitness statistics and hopefully all training data by just going thro...