Changes in Discord to feed AI

[Aleksei]

So its time to forget Discord?
I never liked the thing, as people here might know, but now they made some changes that may be of interest for you all.
Maybe its nothing, but at least its concerning how their ToS changed at the same time that they introduced their AI.

Anyway, I stay faithful to the forums.

If any of you have any more information, please share, and let us know if this is something to worry, or nothing at all =D

The ToS before the change:

Also, is good to mention what @darthmongoose points, about Discord's AI, Clyde.

If you feel the need to change your Discords settings to protect yourself, here are the options that should be off. Thanks to @VibrantFox

[Kelheor]

Aw damn. Well, it's the only place where I really talk to people except for this forum, and we even have our teeny tiny comic community, so I definitely wouldn't want to abandon it. But I guess it could mean less art sharing.

[darthmongoose]

It says that stuff won't be fed into the AI though...

It can only access messages you deliberately send to the AI client, and also the data used by that AI will only be used by that Discord-modding AI, not OpenAI more generally.

Even as somebody who's really not a fan of AI in an art and writing context, I think it's worth reading the announcements and stuff more carefully before jumping to "they're training AI with everything you upload"? The AI in question is a moderation tool.

[Aleksei]

Good! I gonna add to the OP~

[TAMAnnoying]

Altho I agree with you I feel the need to state that this means someone ELSE can feed your art to Clyde.

But like I think people need to understand that this is a general theft problem and not an AI problem- people are dicks and will continue to be dicks.

[VibrantFox]

I saw about this the other day but given how i rarely use the thing anyways I'm not too fussed; tho I did pay attention to a tweet i'd seen about a setting on there about allowing Discord to use your data (somewhere in the privacy section) so going in an checking to make sure that's off is at least one step in kind of protecting yourself. I've got pretty much everything except data for basic functions on but everything else is a big ol no.

Honestly tho the move by discord doesn't surprise me since i remember them (or one of the creators/devs) talking about interest in that web3 whatever before the massive push back they got with people canceling their nitro subs and encouraging others to do the same (maybe folks should do that again) so them being interested in this was like a "disappointed but not surprised" moment

Edit: found the tweets on data use settings
https://twitter.com/Lorcan0c/status/1633965274016653315

[darthmongoose]

I mean.... somebody ELSE can feed your work without your permission to literally any AI... which is one of the many reasons I'm not a fan of it, but also why there are so many lawsuits going on with it right now.

Though it wouldn't change that Clyde is still only meant to use the data purely on discord for moderation and the data won't be sent to feed more general AI. I'm not sure what the purpose would be of adding complex capability to use artworks to generate AI images in a moderation bot? Sounds like a huge waste of money and resources....

Edit to add: I feel like a lot of people don't seem to understand that an AI is a type of program. Not all AI have the same functionality. An AI designed to draw from a database of data for the purpose of modding doesn't necessarily contain the functionality to generate images just because other programs that are AI are written for that purpose.

[Aleksei]

Nice! Gonna add that to the OP too =D

[TAMAnnoying]

I think it might help it determine what "boobs" "dicks" and "butts" are in images because as someone who meets a lot of strangers through a lot of random DND games - lots of people have those "restricted images" where people can't send them porn but MULTIPLE occasions we can send porn to each other that the program doesn't catch and it THINKS things like pretty ladies or the beans from among us are somehow illigal images?

[darthmongoose]

Yes, that's probably what it's intended for.

But people sending it dick pics will likely only be used by this particular AI for the purpose of getting better at recognising dicks, not for the purpose of getting good at creating its own AI generated pictures of dicks, which would be... weird functionality for a mod bot

And thankfully, it's not currently intended to scan ALL images in case they might be dicks or butts; it'll only look at ones you deliberately send to it.

[VibrantFox]

Edit to add: I feel like a lot of people don't seem to understand that an AI is a type of program. Not all AI have the same functionality. An AI designed to draw from a database of data for the purpose of modding doesn't necessarily contain the functionality to generate images just because other programs that are AI are written for that purpose.

I feel this is primarily a consequence of recent happenings with specific kinds of AI hence a lot of fear or general apprehension even at the mere mention of it. I think my bigger issue is more the automatic opting in from certain services to help train these things because even if a person is willing to contribute data there should still be given choice from the start than automatic opt in like a lot of companies seems to be doing lately

---

But on the subject of such a thing being used to automate moderation I still worry for accuracy and effectiveness and what kind of moderation coz lord knows how that's gone on actual media posting sites like tiktok and youtube where auto flagging can get really weird if not straight up stupid

[rajillustration]

My concern is that moderation should just never be done by an AI. Like what's that quote about how a computer should never make a management decision? That and they only clarified that they wouldn't be feeding your calls into Clyde until AFTER twitter had a full meltdown and a bunch of people deleted their nitro, but I'm not that happy with using OpenAI anyway, honestly, it was made through data scraping our words (including alllll of our fanfiction) without consent. I don't care if discord did it or someone else did it. These AI models need to be made ethically.

Also, OpenAI isn't "done", it relies on outsourcing it's data to a human being to make sure the content it has is filtered. Which, if you want to read a little more tea on that, this a good read because like there's way more at stake her than just data. Like it has the appearance of just plug and play and beep boop the robot has a fully formed brain, but no, these AI have to be constantly monitored by a big team of people.

Exclusive: The $2 Per Hour Workers Who Made ChatGPT Safer

[sushy00]

I adjusted my Discord settings now. I hate Discord anyway, I would love to see forums make a comeback!

[system]

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.