Post 1713846 by max@smeap.com
(DIR) More posts by max@smeap.com
(DIR) Post #1710233 by carbontwelve@d20hero.club
2018-12-04T13:25:54.959124Z
1 likes, 4 repeats
This whole #Tumblr debacle just goes to show that AI led community moderation is not only a failure but a dangerous one at that.It would seem that the premise of AI as sold to the general public through mass media is far off into the future and certainly not ready to be making decisions on the behalf of humans,The problem goes further than Tumblr, various governments are beginning to clamp down on social media with the UK having recently made demands that automation be put in place to take down content that is inappropriate citing child safety as the moral panic of choice to get the public on side.Moderators should be thought of as akin to doctors in the real world. In the UK for example there are on average 2.8 doctors for every 1000 people; how many moderators do you think organisations such as Tumblr, Twitter or Facebook have for every 1000 accounts? This is why automation is "key" to their moderating at a cost that enabled their service to be "free." Nobody want to pay for their social media do they?Arguably the Fediverse is a solution to this, with many instances with 10K or less active users, they can often keep the moderators per 1000 users within an acceptable range. However moderation isn't free and in the case of the Fediverse it is entirely donated time kindly given up by volunteers. As instances grow, there may become a point when we have to take a strong look at ourselves and maybe begin crowd funding the Fediverse to keep it a pleasant place to be social within.
(DIR) Post #1710398 by PaladinQuinn@lgbt.io
2018-12-04T13:31:29Z
0 likes, 0 repeats
@carbontwelve how did you do such a long post?
(DIR) Post #1710399 by carbontwelve@d20hero.club
2018-12-04T13:33:06.340226Z
0 likes, 0 repeats
@PaladinQuinn I'm on a #pleroma instance with a 5000 character limit - there are others with higher limits set that are used for blogging.
(DIR) Post #1710580 by maxine@social.mootech.eu
2018-12-04T13:34:32Z
0 likes, 0 repeats
@carbontwelve I agree to latter parts, but the first one is very much a hyperbole.I haven't seen any evidence (I honestly didn't look too, so there's that) that Tumblr uses AI assistance for the newest debacle, but I've seen a lot of evidence for a key search with programmatic mechanism behind, that shall make it look like actual moderationAI are ready though and can make good decisions, even much better than humans do, at current levels.
(DIR) Post #1710581 by maxine@social.mootech.eu
2018-12-04T13:36:46Z
0 likes, 0 repeats
@carbontwelve AI systems are ready, even for moderation, the problem is less them, but lack of understanding of how they work by those who use them.The most important thing for AIs is literally "garbage in, garbage out". They learn exactly what you tell them to learn (by data) and that data is usually garbage.So to sum it up: My point of view here is: yes the system employed are bad, but it's not the system that is bad, but the user who makes it bad (here the corporation).
(DIR) Post #1710582 by carbontwelve@d20hero.club
2018-12-04T13:45:40.624819Z
0 likes, 0 repeats
@maxine I see AI as a force multiplier, if one moderator can take the place of ten with the aid of AI then it is an overall positive contribution. What I hear happen is corporations band about AI like its the lord and saviour of community moderation, with the decisions it makes being final, no human required.I do think that we are in agreement. The systems employed are bad, but it's not the system that is bad, but the corporations implementing them poorly whom makes it bad. I actually wish I had made sure that was more obvious in my long toot, you have better articulated my thoughts haha.Thank you; it's genuinely a pleasure to chat with someone about this who may know more than I do. I love to learn. 😊
(DIR) Post #1710641 by eloisa@masto.pt
2018-12-04T13:35:00Z
1 likes, 0 repeats
@carbontwelve I've got flagged 3 posts. Only one of them actually mine: a photograph inside a History museum with some bones from a big whale. The other 2 were reblogs of feminist stuff. Only one of them containing actually questionable content regarding NSFW.Meanwhile, I do have some artistic nudes in there and so far no flagging.I'm not active in there for ages anymore, but it was my archive. This was just the last push I needed to get out.AI moderation sucks!
(DIR) Post #1710642 by eloisa@masto.pt
2018-12-04T13:35:10Z
1 likes, 0 repeats
@carbontwelve --- Actually, I've been trying to really get out since June this year, but the export tool bugged and I'm being held hostage. Help desk doesn't care.I finally got my data back today thanks to @Nervengift <3
(DIR) Post #1710777 by maxine@social.mootech.eu
2018-12-04T13:56:10Z
1 likes, 0 repeats
@carbontwelve I think the best way to compare it, is if you have an apprentice and then demand from them to make your new flag ship product and expecting them to do extraordinary work.Everyone knows it will fail and we see it. The solution is to put in some other mechanism of quality control, yeah
(DIR) Post #1710779 by Anke@tootplanet.space
2018-12-04T13:52:23Z
1 likes, 0 repeats
@eloisa @carbontwelve bone means boner means erect penis! It makes perfect sense!
(DIR) Post #1710784 by carbontwelve@d20hero.club
2018-12-04T14:00:51.423117Z
0 likes, 0 repeats
@Anke @eloisa 🥁 🙂
(DIR) Post #1710849 by June@kitty.town
2018-12-04T13:58:51Z
1 likes, 0 repeats
@carbontwelve @aparrish if you're not already paying your instance admin either:You're not able, understandableOr:You are able but you don't know your admin or appreciate what they do, in which case you might be on the wrong instance for you.I feel really strongly that if we're gonna make this place work we need to pay admins and mods for their hard work. Literally the only places on the web that feel like a community to me fit this model.
(DIR) Post #1710992 by carbontwelve@d20hero.club
2018-12-04T14:13:30.259104Z
0 likes, 0 repeats
@June You are right, although I would add a third "Or" to your list:Or:Your admin has a small (<100 users) invitation only instance that they keep running because they can afford to do so as a donation to their friends/friends friends whom otherwise cannot.Still in that case I do believe donations should be accepted for when users are able (because in that scenario they would be willing,) to help go towards keeping the whole show on the road is it where.There is a closer feeling of community when people work to build the platform that they use, with the Fediverse that is actually possible and it is nice to see when it happens.Thank you for your reply, it's is genuinely appreciated.@aparrish
(DIR) Post #1711131 by June@kitty.town
2018-12-04T14:17:57Z
1 likes, 0 repeats
@carbontwelve @aparrish sure, basically know your admin and what their needs are. Your scenario is totally great.
(DIR) Post #1712211 by 61@en.osm.town
2018-12-04T14:54:35Z
1 likes, 0 repeats
@carbontwelve#AI most certainly *is* ready to be making decisions on behalf of humans.It is just that the quality of those decisions could do with some improvement. Maybe someone forgot to put that in the requirements specification? 😜
(DIR) Post #1712212 by carbontwelve@d20hero.club
2018-12-04T15:06:22.797431Z
0 likes, 0 repeats
@61 You're not wrong and in reflection I should have made it more clear that I meant in it's current implementation.It appears that AI is being banded about as a cure all by corporations and governments alike rather than as the force-multiplier that it actually is.So instead of AI being used to flag things to a human it appear to be that all AI decisions are final no human required and it is that, which I believe to be dangerous.Until we have C3P0 level of AI intelligence I don't think we should be taking humans out of the loop, at least not for now.
(DIR) Post #1713001 by ao@p.a3.pm
2018-12-04T15:38:06.617073Z
1 likes, 0 repeats
@carbontwelve counterpoint: leaving moderation to humans is bad and scarring
(DIR) Post #1713002 by ao@p.a3.pm
2018-12-04T15:38:49.161008Z
1 likes, 0 repeats
@carbontwelve not saying that AI moderstion is good, but human moderation is not without problems
(DIR) Post #1713063 by carbontwelve@d20hero.club
2018-12-04T15:43:46.101827Z
0 likes, 0 repeats
@ao oh for sure, I glossed over a lot of the topic to keep the toot a reasonable length; in hind sight I wish I had made it more clear that I am not against AI, just the current miss-advertising of it as a cure all by corporations and governments because in current forms its often worse than the human moderation it replaces.I do believe AI has a role to play in moderation, just the current trend of AI having the last say, no human required is dangerous. AI should be a human force multiplier not a human replacement. We want better fairer moderation not just different moderation.
(DIR) Post #1713200 by carbontwelve@d20hero.club
2018-12-04T15:49:53.133410Z
0 likes, 0 repeats
@ao it's a complicated topic for sure and actually having thought a lot about it today it's one that I think I am going to invest some time researching and writing about in the new year.I have had some interesting, thought provoking replies, including yours and it's important that I hear ideas from points of view other than my own, thank you for your replies they are very much appreciated. 🚀
(DIR) Post #1713846 by max@smeap.com
2018-12-04T16:10:11Z
1 likes, 0 repeats
@carbontwelve I think you are right, we need to being people back into the equation before it is too late. Lately I've been pointing people to this talk to explain how broken our understanding of AI/ML currently is, because it's funnier than just listening to me doom and gloom about it: https://www.usenix.org/conference/usenixsecurity18/presentation/mickens
(DIR) Post #1714922 by efflorescence@sunbeam.city
2018-12-04T16:56:16Z
1 likes, 0 repeats
@carbontwelve i really like this take, but would you mind cwing posts this long as long or something? i feel like it helps keep down the clutter.tysm :blobheart:
(DIR) Post #1714923 by carbontwelve@d20hero.club
2018-12-04T16:59:33.300677Z
0 likes, 0 repeats
@efflorescence apologies for that.
(DIR) Post #1715041 by carbontwelve@d20hero.club
2018-12-04T17:04:11.222779Z
0 likes, 0 repeats
@efflorescence (I hit toot before I finished) I forgot there are some clients that don't do read more on the long toots. I will remember to cw in future. Thank you.
(DIR) Post #1719558 by mindshoot@mastodon.xyz
2018-12-04T19:05:36Z
1 likes, 0 repeats
@carbontwelveThis toot, and the civilised discussion following it, really captured what it is I like about the #fediverse!
(DIR) Post #1719559 by carbontwelve@d20hero.club
2018-12-04T19:50:32.256215Z
0 likes, 0 repeats
@mindshoot this is generally the feeling I get from Mastodon.I'm not sure if its because I have cultivated a group of followers/following that are largely civilised, respectful and willing to both be patient when they feel someone is wrong so they can articulate their feelings without conflict and open to learning something new when the opportunity presents.Or it could be that a majority of Fediverse (sans some defederated instances) is like that. I have a sneaking suspicion that it _is_ the later; aside from the handful of obvious trolls everyone I have interacted with on this platform have been super nice.
(DIR) Post #1720027 by 61@en.osm.town
2018-12-04T19:45:35Z
1 likes, 0 repeats
@carbontwelveYeah I understood what you meant and was just inflicting my sense of humour upon you. Quite unprovoked, apologies for that. 🙂That said, you make a pertinent observation on the current abuse of #AI tools and solutions.This, shall we say, blind optimism seems to be a general trend coming from silly valley #VC backed startups. Which I suppose is OK until you become persuaded that future potential = present performance.Remember #theranos? Textbook example.
(DIR) Post #1720028 by carbontwelve@d20hero.club
2018-12-04T20:02:18.391700Z
0 likes, 0 repeats
@61 that's fine; I thought you were being funny and there was no offence taken you made me smile.I wonder if there is a current line of academic investigation into the use of AI to enforce behaviours in humans. I'd love to write a book on the subject but alas I have bills to pay 🙄 #theranos is indeed a textbook example, as I believe is the recent (June?) scaling back of IBM Watson Health (#Phytel ?) due to it not working; although I believe that was more due to IBMs mismanagement of the software and going for profit over getting it to work.
(DIR) Post #1723310 by mindshoot@mastodon.xyz
2018-12-04T21:45:52Z
1 likes, 0 repeats
@carbontwelveIn my more optimistic moments I have a feeling that, in fact, most people on the planet are generally reasonable and decent. It is just unfortunate that the current model of the internet in general favours noise and extreme views above sense. Perhaps the breadth of the fediverse community encourages tolerance; long may it last!
(DIR) Post #1962937 by pootz@mastodon.technology
2018-12-13T19:02:55Z
1 likes, 1 repeats
@carbontwelveThe internet needs to be ran on publically funded server space and moderation should be democratized so that community members are entrusted with some level of moderation as well.X # of users report a post on the local timeline? Hidden until appealed.Mods can decide if it's abuse and revoke that if necessary, but ultimately the community should be responsible for its own moderation.Because nobody is going pay to use the fediverse if corporations can do it for free.
(DIR) Post #1966207 by Shamar@mastodon.social
2018-12-13T21:05:27Z
0 likes, 1 repeats
@carbontwelve @ao As one who was banned by a couple of technical communities with dumb excuses for asking the wrong questions or challenging the wrong assumptions (see https://dev.to/shamar/i-have-been-banned-from-lobsters-ask-me-anything-5041 and http://www.tesio.it/cache/2018/TildesNET_Undetectable-Remote-Arbitrary-Code-Execution-Attacks-through-JavaScript-and-HTTP-headers-trickery.html if you are interested in the details or you work in #InfoSec), I still prefer human moderation.Because when people abuse their power they will be held accountable (in the long term).With machines nobody would respond of convenient software's errors.
(DIR) Post #1966215 by Shamar@mastodon.social
2018-12-13T20:56:37Z
0 likes, 1 repeats
@carbontwelve @61 #Computers compute.They do not decide.http://www.tesio.it/2018/10/06/the-intelligent-symbiosis.html
(DIR) Post #1968929 by bob@soc.freedombone.net
2018-12-13T21:17:54.568540Z
0 likes, 0 repeats
@61 @carbontwelve I did a lot of AI, or AI-adjacent, things in the past and I'm not so confident that AI can be used for community moderation or governance of communities. I expect the Facebook push for AI will fail to deliver what they expect.It might be useful in situations where the problem can be exactly specified and is context free, but in open-ended conversations on social networks even the best AIs today aren't going to understand the context and are going to make bad decisions.
(DIR) Post #1968930 by maiyannah@community.highlandarrow.com
2018-12-14T00:01:42+00:00
0 likes, 0 repeats
@bob @61 @carbontwelve It isn't being used for actual moderation, though. It's being used to diffuse responsibility for censorship.
(DIR) Post #1977779 by linkskywalker@soc.ialis.me
2018-12-13T21:37:41Z
1 likes, 0 repeats
@June @carbontwelve @aparrish My first reaction to reading this was "Absolutely! Though my instance doesn't accept donations." Then I got curious. Then I did some digging. Turns out my instance DOES accept donations, I just never realized it. Thank you for reminding me to live up to my beliefs.
(DIR) Post #1978207 by carbontwelve@d20hero.club
2018-12-14T09:22:21.026263Z
0 likes, 1 repeats
@bob I think I said similar in a previous thread but for auto moderation of video and/or photos AI can be a great force multiplier to a human moderator, actively acting as a bloom filter on content so that the human can make better decisions or have content that would otherwise go under the radar get double checked by a human.They _are_ terrible for anything contextual and until they get a level of intelligence equivalent to C3P0 I would not trust AI to moderate solo.The danger I alluded to in my original toot was not that AI is dangerous (it can be) but that the trend of Governments mandating AI use and corporations advertising AI as a cure all (snake oil) is.@61