[00:27:55] (03PS10) 10Krinkle: Introduce ext.ores.api [extensions/ORES] - 10https://gerrit.wikimedia.org/r/459549 (https://phabricator.wikimedia.org/T201691) (owner: 10Ladsgroup) [00:28:40] (03PS3) 10Krinkle: ext.ores.api: Single function [extensions/ORES] - 10https://gerrit.wikimedia.org/r/483338 [00:29:05] (03CR) 10Krinkle: [C: 03+2] "Fixed eslint issue that was from a rule recently enabled in master from another commit." [extensions/ORES] - 10https://gerrit.wikimedia.org/r/459549 (https://phabricator.wikimedia.org/T201691) (owner: 10Ladsgroup) [00:33:49] (03Merged) 10jenkins-bot: Introduce ext.ores.api [extensions/ORES] - 10https://gerrit.wikimedia.org/r/459549 (https://phabricator.wikimedia.org/T201691) (owner: 10Ladsgroup) [00:33:52] (03Merged) 10jenkins-bot: ext.ores.api: Single function [extensions/ORES] - 10https://gerrit.wikimedia.org/r/483338 (owner: 10Krinkle) [00:35:55] (03CR) 10jenkins-bot: Introduce ext.ores.api [extensions/ORES] - 10https://gerrit.wikimedia.org/r/459549 (https://phabricator.wikimedia.org/T201691) (owner: 10Ladsgroup) [00:37:52] (03CR) 10jenkins-bot: ext.ores.api: Single function [extensions/ORES] - 10https://gerrit.wikimedia.org/r/483338 (owner: 10Krinkle) [00:59:30] (03PS1) 10Krinkle: build: Structure updates (and upgrade eslint config) [extensions/ORES] - 10https://gerrit.wikimedia.org/r/485138 [00:59:32] (03PS1) 10Krinkle: ext.ores.highlighter: Fix minor memory alloc [extensions/ORES] - 10https://gerrit.wikimedia.org/r/485139 [01:01:24] (03PS2) 10Krinkle: build: Structure updates (and upgrade eslint config) [extensions/ORES] - 10https://gerrit.wikimedia.org/r/485138 [01:01:26] (03PS2) 10Krinkle: ext.ores.highlighter: Fix minor memory alloc [extensions/ORES] - 10https://gerrit.wikimedia.org/r/485139 [11:10:31] (03CR) 10Ladsgroup: [C: 03+2] build: Structure updates (and upgrade eslint config) [extensions/ORES] - 10https://gerrit.wikimedia.org/r/485138 (owner: 10Krinkle) [11:16:05] (03Merged) 10jenkins-bot: build: Structure updates (and upgrade eslint config) [extensions/ORES] - 10https://gerrit.wikimedia.org/r/485138 (owner: 10Krinkle) [11:18:20] (03CR) 10jenkins-bot: build: Structure updates (and upgrade eslint config) [extensions/ORES] - 10https://gerrit.wikimedia.org/r/485138 (owner: 10Krinkle) [11:27:56] 10MediaWiki-extensions-ORES, 10Scoring-platform-team (Current), 10MW-1.33-notes (1.33.0-wmf.14; 2019-01-22), 10Patch-For-Review, 10User-Ladsgroup: Implement JS ORES client in mw-ORES extension - https://phabricator.wikimedia.org/T201691 (10Ladsgroup) @Halfak This task is for a basic implementation of the... [12:08:05] I forgot to say o/ [12:30:28] 10ORES, 10Scoring-platform-team (Current), 10Documentation, 10User-Ladsgroup: Document/script ORES config change deployment process - https://phabricator.wikimedia.org/T213743 (10Ladsgroup) a:03Ladsgroup https://wikitech.wikimedia.org/wiki/ORES/Deployment#Unusual_maintenance_actions [14:03:04] (03CR) 10Esanders: ext.ores.highlighter: Fix minor memory alloc (031 comment) [extensions/ORES] - 10https://gerrit.wikimedia.org/r/485139 (owner: 10Krinkle) [14:46:48] 10ORES, 10Scoring-platform-team (Current), 10Patch-For-Review, 10User-Ladsgroup: Change default serializer of celery from pickle to json - https://phabricator.wikimedia.org/T206333 (10Ladsgroup) done now [14:55:04] 10Scoring-platform-team, 10Release Pipeline, 10Wikibase-Containers, 10Wikidata, and 3 others: Stretch in docker registry forces ascii encoding - https://phabricator.wikimedia.org/T210260 (10Ladsgroup) >>! In T210260#4823617, @hashar wrote: > @Ladsgroup if you retry blubber for wikilabs, you should no more... [15:05:16] hey folks. I'm going to be a bit late today. I have to pedal to the university and I have meeting stating riht when I get there. [15:05:25] o/ Amir1 & harej ^ [15:05:32] I should be back on IRC in about 2 hours. [15:05:52] no worries. I'll be gone by then [15:09:19] Amir1, can you update current work before oyu head out? [15:09:34] sure [15:09:44] I see you updated it already for your work earlier this week :P [15:09:48] OK I'm off [15:09:49] o/ [15:24:11] OK turns out I didn't have my bike ready to pedal in and it would take too long to adjust (curse the changing weather) so I'm gonna be skipping going to the U today. [15:30:45] That means I'm available for at least the next 30 minutes. [16:34:25] Yup. I'm just going to not go in. I'm sitting at my computer in my bike tights. :| [16:43:56] Amir1, if you're still around, I'm going to push a change to the revscoring json branch. [17:48:22] (03PS1) 10Umherirrender: Use tabs in extension.json [extensions/ORES] - 10https://gerrit.wikimedia.org/r/485223 [17:48:44] ... I don't even see why we need a timestamp_str. [17:48:59] It looks like timestamp comes from a lower dependency on rev_doc. [17:52:57] I'm looking in ORES to try to work out how we gather base dependencies. [17:58:08] halfak: have you ever met John Bohannon? [17:58:30] Doesn't ring a bell. But I might have. Where might I have met him? [17:59:04] https://en.wikipedia.org/wiki/John_Bohannon [17:59:08] That guy? [17:59:29] Oh cool! He's the guy who did the Dance your PhD stuff. [17:59:32] Yes [18:01:21] Aha! I see that "trim" will get down to the revision_oriented datasources. We could instead do a "trim" with the extractor context to get down to the rev_doc. [18:02:44] Oh wait. I was wrong. We use "dig" but again it's context free. [18:02:53] We instead want to use a context-based "dig" [18:04:53] Yus! I think this will work. But is it the best way? [18:05:11] Essentually, we'll be passing around rev_docs and user_docs returned by the mwapi [18:05:46] The biggest harm that I can see is that we'll be passing slightly more information to a celery worker than before because there are some parts of a rev_doc that we don't use. [18:05:55] Probably less than 2% of the data. [18:06:26] Also, it would be good to have all of our "revision_oriented" datasources be simple datatypes. [18:09:18] OK yeah this doesn't work either. [18:09:23] * halfak works more. [18:11:23] (03CR) 10Awight: [C: 03+2] Use tabs in extension.json [extensions/ORES] - 10https://gerrit.wikimedia.org/r/485223 (owner: 10Umherirrender) [18:16:02] (03Merged) 10jenkins-bot: Use tabs in extension.json [extensions/ORES] - 10https://gerrit.wikimedia.org/r/485223 (owner: 10Umherirrender) [18:18:46] (03CR) 10jenkins-bot: Use tabs in extension.json [extensions/ORES] - 10https://gerrit.wikimedia.org/r/485223 (owner: 10Umherirrender) [18:38:16] 10Jade, 10Scoring-platform-team (Current), 10DBA, 10Operations, and 3 others: Introduce a new namespace for collaborative judgements about wiki entities - https://phabricator.wikimedia.org/T200297 (10awight) >>! In T200297#4887041, @Krinkle wrote: > 3. The feature proposes to store arbitrary text (specific... [19:25:12] Lunch! [20:29:45] Ultimately, I made some minor changes to timestamp behavior and pushed a commit. [20:37:00] 10Scoring-platform-team, 10artificial-intelligence: Implement NSFW image classifier using Open NSFW - https://phabricator.wikimedia.org/T214201 (10Halfak) [20:52:23] harej, ^ now we're getting into the interesting stuff. I don't know how we could manage a labeling campaign for this :| [21:01:38] Unrelated. I finished the add proposal/endorsement wireframes: https://docs.google.com/drawings/d/1urHCaEycUs0n-63bTcWLOJtaRE7D08yWd8nQS6W6Jd0/edit [21:15:04] I think you meant to refer to something but remember I have the spam bots on ignore [21:15:47] I'm confused. [21:15:57] I referred to a phab task for catching NSFW images. [21:16:21] Oh! You don't see wikibugs! [21:16:27] Bingo. [21:16:28] Implement NSFW image classifier using Open NSFW - https://phabricator.wikimedia.org/T214201 (Halfak) [21:16:41] Ooh, this will be interesting [21:16:54] Occasionally there are also calls for Commons to implement its own version of Content ID [21:17:03] For identifying copyright-protected works. [21:18:03] Right. Probably not a business we want to get into. [21:21:11] I'm not sure that Commons cares about a NSFW scanner per se. [21:21:20] They're... pretty open to adult content [21:21:35] As long as there's some pretense of artistic or educational value. [21:21:44] What they *don't* want is *low quality* NSFW content. [21:21:50] Or selfies. [21:31:50] harej, we have a specific request. It's not about removing NSFW content from commons, but flagging when NSFW content is added to an article -- for review. [21:32:17] Ooh, that's interesting. Do you know more about the use case? [21:32:57] I'll CC you on the thread :) [22:08:35] 10Scoring-platform-team, 10artificial-intelligence: Implement NSFW image classifier using Open NSFW - https://phabricator.wikimedia.org/T214201 (10Krenair) Given previous controversy around this sort of thing, we should be extremely careful about how such a classifier gets used. [22:13:58] 10Scoring-platform-team, 10artificial-intelligence: Implement NSFW image classifier using Open NSFW - https://phabricator.wikimedia.org/T214201 (10Krenair) #wikimedia-ai logs: 2019-01-18 21:31:50 harej, we have a specific request. It's not about removing NSFW content from commons, but flagging when... [22:16:01] 10Scoring-platform-team, 10Release Pipeline, 10Wikibase-Containers, 10Wikidata, and 3 others: Stretch in docker registry forces ascii encoding - https://phabricator.wikimedia.org/T210260 (10greg) 05Open→03Resolved >>! In T210260#4892443, @Ladsgroup wrote: >>>! In T210260#4823617, @hashar wrote: >> @Lad... [22:16:30] 10Scoring-platform-team, 10artificial-intelligence: Implement NSFW image classifier using Open NSFW - https://phabricator.wikimedia.org/T214201 (10Halfak) Uh. That's IRC and this is a new task :) Thank you for copying them though. [22:21:01] (03CR) 10jenkins-bot: Localisation updates from https://translatewiki.net. [extensions/JADE] - 10https://gerrit.wikimedia.org/r/485292 (owner: 10L10n-bot) [22:22:25] 10Scoring-platform-team, 10artificial-intelligence: Implement NSFW image classifier using Open NSFW - https://phabricator.wikimedia.org/T214201 (10Krenair) Yes. I believe that when opening a task around a subject that may involve controversy, it can be helpful to provide any context showing how this is not the... [22:23:30] Krenair, maybe we could discussing things here. [22:24:17] I'm in a meeting right now, and I created a placeholder task so that I could come back to it later. [22:24:22] Please just wait a few minutes. [22:27:49] Sure, I'm not in any hurry [22:31:46] OKAY! Done with meeting. [22:32:13] So, first, Krenair what's going on, man. I don't see anywhere in that task that I said we were going to do anything controversial. [22:32:18] okay [22:32:40] I'm not here to trash your idea or project [22:32:48] Is having a model that could tag images as NSFW controversial? [22:32:51] I'm not here to rush you into forwarding everything into the ticket [22:33:07] I am quite happy to be patient [22:33:25] I am interested in this project [22:33:47] That was a lot of posts on the ticket in quick succession -- and a lot of being called out. [22:34:12] Honestly, I put that task into place quickly while switching between tasks to make sure that the conversation about it moves to phab. [22:35:01] I'm not calling you out [22:35:16] I don't think you've done anything wrong [22:35:31] I'm sorry if it appeared that way [22:36:28] 10Scoring-platform-team, 10artificial-intelligence: Implement NSFW image classifier using Open NSFW - https://phabricator.wikimedia.org/T214201 (10Halfak) [22:36:49] I just added some details about the proposal. [22:37:00] yeah [22:37:50] I would be worried when making such a ticket about what assumptions people may have [22:38:14] Seems lots of people could have lots of assumptions about things. Honestly, I don't know what your assumptions are. [22:38:22] sure [22:39:18] I personally believe having a filter that can tag images as NSFW is fine [22:39:26] I suspect some others disagree [22:39:34] The use case seems great [22:39:56] Oh? What types of disagreements do you see? [22:40:25] Seen several incidents of vandalism violating least astonishment [22:40:27] well [22:40:57] Some people may be worried that this could make it easier to censor content in future [22:41:40] Like removing all NSFW content from commons or something like that? [22:41:55] possibly, or just another step towards the personal image filter [22:43:09] I don't think you're looking to do that [22:44:22] 10Scoring-platform-team, 10artificial-intelligence: Implement NSFW image classifier using Open NSFW - https://phabricator.wikimedia.org/T214201 (10Halfak) [22:44:25] Oh yeah. I totally forgot about image filter. [22:44:32] right [22:44:32] I just added some more background to the task. [22:44:40] you registered before then [22:44:59] 10Scoring-platform-team, 10artificial-intelligence: Implement NSFW image classifier using Open NSFW - https://phabricator.wikimedia.org/T214201 (10Halfak) [22:45:00] I actually wasn't around at the time but have since reviewed some snippets [22:45:03] Oh yeah. I've been around for a while :) [22:45:46] to get communication around this wrong is a potential political landmine [22:46:45] Wikimedians treat their porn the way that the american south treat their guns [22:46:53] lol [22:46:53] I felt I was doing you a minor favour by posting your message clarifying the use case here to the task [22:46:54] 10Scoring-platform-team, 10artificial-intelligence: Implement NSFW image classifier using Open NSFW - https://phabricator.wikimedia.org/T214201 (10Halfak) I added some details to the task description that should at least make the intention clear. This task is still very much in our "backlog" and it's sitting... [22:47:04] I knew the link to the task had already spread [22:47:10] Krenair, indeed I did appreciate you posting that detail. [22:47:33] It was the conclusion that harej jumped to as well so it's good to clarify. [22:47:42] I do like transparency though :p [22:48:02] though you arguably have enough here already [22:49:30] FWIW, I literally read the email related to the task 5 minutes before creating a task. This is as transparent as I get :P [22:49:44] heh okay [22:50:04] I didn't want to wait to create the task and let the email thread go on so I figured a quick task stub was a good idea. I can see now that it has the potential to confuse. [22:50:33] that's often a good idea, in cases such as these I'd say it could be a risk [22:51:22] IIRC, parts of discussions around this were about who gets to decide what is NSFW and what is not [22:51:38] Heh. In this case, it's the vandals! [22:51:52] with I think the general assumption being that a human patroller has to do it? [22:52:13] Right. I can't imagine something like this working automatically. [22:52:30] that's a pretty big deal when it comes to deciding who gets to see what [22:52:44] in this case with an AI-powered AbuseFilter-like tagging thing [22:52:47] it's probably not a big deal [22:52:51] or at least I think it isn't [22:53:18] Right. change tags could go a long way. At least with abuse filter it would be under direct control of editors with appropriate rights. [22:53:22] OK on that note, I'm going to hit the road. Have a good weekend folks! [22:53:26] okay [22:53:27] And thanks for your help, Krenair [22:53:29] o/ [22:53:35] have a good weekend [23:01:41] halAFK: if "create a public API to identify NSFW [or otherwise controversial] images" ends up on our roadmap, I would like to do some research into the image filter fiasco from several years ago, if nothing else so we can (a) develop our API in a particular way in response (b) figure out best practices for API usage (c) managing messaging around the API and its purpose [23:02:39] One potential use of such an API is that someone will be able to develop the dreaded "porn filter" as a gadget. Depending on the wiki, that gadget could be turned on by default. [23:03:24] yeah that's pretty much what I had in mind when I suggested it might be seen as "just another step towards the personal image filter" [23:03:43] unlikely to happen on en, maybe [23:04:24] as we've seen recently, some more obscure wikis might have differing ideas [23:05:24] if this API checks strictly for nudes and things like that, then the problem is limited in scope. if this is general image recognition (which is still a good idea, just more complicated), then not only do we get porn filters, we also get Muhammad filters and so on. [23:06:06] Keep in mind, that mere proposals like "Lets put all the nsfw images in a category" are insanely controversial [23:06:29] TBF what is not safe for work is pretty subjective [23:06:50] I'm not sure if taking the human out of the equation will make it more or less likely to be controversial [23:07:22] yeah what is safe for work on arwiki is likely not the same as what is safe for work on enwiki :) [23:08:35] either way it's worth considering exactly how big a political landmine this really is [23:10:33] I would want to work with CE to float some kind of trial balloon to get a sense of what communities thought of the general concept, ideally at the high level without talking about specific use cases