#but asking AI questions instead of looking it up using chat gpt to write AI generated filters and headcanons and all that

[marbleheavy]

i actually do think i’m better than people who use AI

[1989tv]

Hey, Katie here, we've interacted a few times. Idk if you remember but we briefly talked about ai stuff and ethical uses.

Idk if you've seen any of my posts on the topic, but I have a real (educated comp sci) concern about the things that are happening ongoing right now re: ownership of digital rights, and I know you mentioned working in or around the industry?

I am trying to help avoid a situation where rich techbros or old white guys end up with the power to keep replicating and replacing real artists. I'm also trying to protect swifties and other Fandom from prolific fraud.

I have been ghosted by nothing but bots and voice-mail, and this tech is changing very fast, writers and programmers and traditional artists are already being replaced. They've been doing it with actors too without their consent. I have a real solution, but cannot get through to anyone.

What should I do? Feel free to answer privately, I'm asking as a professional, not a "fan"

hey katie thanks for this question! it’s all really important and there’s lots here and related to this topic that need to be considered and evaluated (even beyond related to artists). this turned out longer than i expected so i’ve put this under the cut so people can ignore if they do choose

so i don’t know how much of an answer i’ll have for you but i can give some background to it. so as of now there really aren’t any laws when it comes to AI or software in general for that matter. there’s never been a precedent for it because of that no reason for laws to go into place. i will say certain regions have laws that protect privacy and data more strictly but again not really any laws to prevent things that could go wrong or the unethical usage of software. usually when someone gets in trouble or taken to court due to software it usually (in the past) has been because it’s had an overlap for some other crime.

so a lot of these companies that are working with AI rn do practice ethically but this is to an extent. it’s to ensure their own fine tuned versions of the models don’t produce anything unethical (so you can’t ask bings chat gpt or google bard to sing and produce a new taylor swift song. it might write a song but the audio won’t occur) a lot of content does filter out. additionally while they might be practicing ethical ai it’s mainly to ensure their users and customers have a guarantee that their prompts and data will be secure and used ethically when further fine tuning the model. they aren’t out here fighting unethical uses of these services. a lot of times especially when it involves malicious intent these wrong uses of AI will be coming from private groups or hackers (not always the case however). so a lot of these companies probably aren’t responding bc of that. the big hurdle is getting big companies to really push for and start a precedent of this which will be….hard to say the least.

im sorry to hear these companies you’ve reached out to are doing this. id really encourage you to keep trying and instead of passing along your idea maybe start with asking them steps they are taking to ensure they are using ethically. more than likely they’ll be happy to share. try reaching out to hr or security departments directly if this is available. aside from that i think one of the best ways to really get your idea and concerns out is to reach out to companies that are actively trying to fight this and we’re specifically created with this in mind (or start your own! a lot of work and money i know 😵‍💫 and not optimal in any way) a lot of these will be new or start up companies. this wave of AI is new and we’re heading into some uncharted territories with it. so while there are people using this wrongly there are also people with the same idea as you in mind and looking to protect their well being. i think when done correctly this can be really exciting and life changing for lots of people (like i said when used correctly) but i agree 100% with you on the dangers of unethical usages of AI. this goes beyond just celebrities or actors. and while it’s not something that’s new (think deepfakes) it’s certainly now more heightened and the risk of everyone could be involved

[enbeees]

You know Chat-GPT is SO fun (and let me tell you right now, getting it to write a silly story or script with existing characters is so fun bc it gets personalities SPOT ON) but it genuinely worries me that so many people seem to trust it implicitly to the point of even admitting to using it instead of Google or whatever (idk maybe its all tech bros but I see this uncomfortably often)

Like oh my God do NOT do that! Aside from problems with memory in long conversations or games, it WILL make up answers and create fake sources for them! Not to mention that the guidelines and censorship mean you may not get the full accurate answer in the first place.

Like I've tested it with Chat-GPT and You.com (which i believe uses the same technology) and as far as I could tell, they purposely skirt around "uncomfortable" topics if they were in an answer, or outright give incorrect information to avoid them.

Tests below readmore. They include spoilers for Fate/Zero!

Now there's a BUNCH of incorrect info in these, which is obvious if you've seen the show.

Chat-GPT:

- Risei did not provide Kirei with Gilgamesh

- While Gilgamesh and Kirei do have some differing views and Kirei doesn't particularly like him at first, they ultimately form a partnership because they're so similar

- Kirei's fate was definitely not left open-ended.

- The exact details of how their partnership was formed were shown in the anime. They were a major plot point, even! We saw basically every detail!

- Kirei did not manipulate Gilgamesh into working with him, Gilgamesh essentially manipulated Kirei into working with him.

- Do I even have to mention that yes, Kirei stole Gilgamesh from Tokiomi

- Kirei does not "encounter Gilgamesh during a battle," he's been working with Tokiomi and Gilgamesh seeks him out to hang with him because he finds him interesting, and Gilgamesh heavily suggests they form a partnership

- By the time Kirei has teamed up with Gilgamesh, Gilgamesh and Tokiomi are very much not still Master and Servant

You.com chat:

- Right off the bat it fucked up because apparently Kirei has now stolen Illya's servant at some point?? He is not and has never been Berserker's master

- Kirei isn't really known for his magecraft?? I mean I guess I can get this bit because he is exceptional at it but he doesn't focus on it and that's the whole reason he had to train alongside tokiomi before the war

- I'm not touching the "his goals are related to the origin of the grail itself" i mean,,, in a way, i guess??? But not really??

- Kirei and Gilgamesh do not have a complicated alliance fraught with tension and conflict,,, those two team up and just have the fuckin time of their lives for the rest of the show (also kirei is very much not gilgamesh's advisor)

- Again, they didn't form a partnership without Tokiomi's knowledge (well, in the beginning they started to), he was dead by the time Gil became Kirei's servant

- Do I even have to say anything about the fact that it claims Kirei never did anything to Tokiomi or acted against him in any way. Especially the "there is no direct conflict between them that results in harm being done to Tokiomi by Kirei." bruh

Like okay, we've cemented the fact that it gives inaccurate information quite often.

But I assume due to its own bias and guidelines in being wholesome and positive all the time, if you've watched the show, you'll realize there is a major, major plot point that it seems to continuously skirt around, leave out, or even lie about (looking at you you.com).

Kirei fucking killed Tokiomi!

I did manage to get chat-GPT to admit it with "what did kirei do to tokiomi" but you.com, as you can see, straight up lied when I asked that question.

So like idk maybe don't consider an ai trustworthy when it comes to accurate information, especially when said ai will skirt the truth and leave out important details to avoid talking about "dark" content, and especially when it is often just outright wrong?