|
I tried using chatGTP a week or so ago to do some improvisational roleplaying and while it insisted on reminding me constantly that we were just doing a hypothetical, it definitely remembered what we were talking about throughout the conversation. And the conversation is still available to continue, if I want. I think the algorithm probably goes back through the chatlog and references things already discussed, so it might just be a repetition thing, and it may or may not correctly interpret previous things both it and I said. So it probably gets facts backwards or mixes things up sometimes and that's not necessarily due to a lack of "memory" so much as the algorithm not being good enough at parsing its own output for key facts and how they change over the course of the conversation. Here's what we said: https://pastebin.com/pFAmn3zC BrainDance posted:That's why it's gotta be open source and able to run on consumer gpus. So I can have my AI that I just teach about SNES rpgs and the Nazis can go have theirs and not bother me. Those sure are some scarequotes 
|
#
¿
Jan 30, 2023 17:13
|
|
|
|
| # ¿ May 22, 2024 09:27 |
|
|
Doctor Zero posted:To me, ChatGPT seems so eager ( Yeah, it's obviously really primitive in what it's doing with this format right now, and disagreeing with the user shouldn't always be seen as a failure mode. But it is "remembering" at least in some respect, I think by re-crawling/indexing what's already in that chat when it generates new output.
|
|
#
¿
Jan 30, 2023 21:04
|
|
) to try to give the human what they want that it ends up contradicting itself or changing its assumptions (
|
drat well I started by like, hmm I should look into this tool, and this next tool, and... but where I wound up is "this person already has a lot of skills I don't" so I'm just spectating now. But it's extremely cool to see someone engage with the technology and apply artistic choices in a conscious way to produce a result. I see this as not much different from an artist I know who works in collage - any kid can cut things out of magazines and make a collage, but raising it to an art form means using the ephemera the world gives you and then applying a combination of choices (artistic vision, whatever you want to call it) and technique (skill with the tools). Only by having both can you wind up with good art. I'm really impressed, Kaker, and I hope you keep at it.
|
|
#
¿
Feb 9, 2023 19:59
|
|
|
The legal hazard of publishing AI art (as opposed to using it for noncommercial purposes) could be significant. Even leaving aside the ethical arguments, if you are contemplating publishing something for money or even just for cred, you might want to not put AI art in it. Unrelated: The most recent Last Week Tonight with John Oliver was focused in its main story on AI art and it's not a bad watch, even if it glosses over some of the technical details a bit. Leperflesh fucked around with this message at 23:53 on Mar 1, 2023 |
|
#
¿
Mar 1, 2023 23:51
|
|
|
Boba Pearl posted:I don't understand how this doesn't lead to people just lying? If Paizo says "you're not allowed to put AI generated art in stuff" and then people lie about it, that absolves Paizo of legal responsibility. It's cover. Again they may actually also be taking a moral stand, but I feel like the legal cover part is the important one that their lawyers may have been involved in. Actually it's also moral cover. If someone says "hey, such-and-such content on your site contains images that were clearly an AI ripping off my art/my client's art" there's less chance of a brigade of outraged internet people going "woah, Paizo supports plagiarism!" if Paizo can point to their policy and say "no we don't, we specifically said this isn't allowed, this is one person's fault and of course we're taking down the art immediately" etc. Leperflesh fucked around with this message at 00:04 on Mar 2, 2023 |
|
#
¿
Mar 2, 2023 00:01
|
|
|
The big money who make money from advertising on their websites that people don't need to visit if their content has been scraped and is being regurgitated by the new chat AI frontends to search, presumably. Similarly to why Getty Images is going after image generators. AIs paraphrasing your content means you get no clicks. It's not even telling you sources unless you explicitly ask for them. The natural reaction if there is no regulation will be to pull everything of value behind paywalls to prevent the AIs from being able to scrape them, and that's not a great result for people interested in continuing to have all that free content. Robot.txt isn't permission to copy your content, it's to index it. This is a new case in which "indexing" is allowing an offsite company to reproduce your content, paraphrased and blended with other site's similar content. The people that own these AI programs are trying to make the argument that it's the same thing, and that argument is probably going to wind up having to be tested in court. The practical outcome will be that people who can, will put up paywalls to prevent a big chunk of their traffic from being redirected to an AI frontend they have no control over and get no revenue from. People who can't, may just take down their stuff, or find avenues to share it where it can't be scraped.
|
|
#
¿
Mar 17, 2023 16:32
|
|
|
I'd like to invite folks to have a discussion about moderation of AI topics in TG, over in the stickied feedback thread. This can include whether/where we could have a space for discussion of AI-generated content and projects without general criticism of AI, or if that's just not a thing that should exist, etc. Meanwhile, I'll continue to let conversations run here, and encourage people engaging with these tools for their projects to continue to talk about them. Trolling still isn't allowed, per general TG rules, and I'll ask even folks who are adamantly against AI tools to please be respectful.
|
|
#
¿
Mar 20, 2023 19:27
|
|
|
There's a report and some frustration because this is a thread for AI tools for trad games, but what I said earlier was that debate & discussion is still allowed provided it remains civil. There is no other place in TG to have this conversation, basically, and the more generalized conversations in D&D and maybe GBS aren't about AI works in trad games. That said: there are some intractable positions here that clearly are not going to get resolved, so at some point folks, you'll need to recognize that you've made your point.
|
|
#
¿
Mar 24, 2023 15:48
|
|
|
PlaysGamesWrong posted:This is objectively incorrect. Yeah just resorting to insults will result in probations, but I like to make the limits clear first. Ghost Leviathan posted:That's just because you're a wuss. This crosses the line.
|
|
#
¿
Mar 24, 2023 23:43
|
|
|
PlaysGamesWrong posted:Thread is just full of garbage people now. Unbookmark. Nah. The training of AIs on copyrighted content without permission is problematic and it'd be gross to disallow anyone from saying so. I have been playing with these tools myself, I'm currently on the side of "for private use this is basically OK, although the AI companies need to compensate the content owners they're training their tools on" but I don't get to impose my personal ethical stances on the whole forum. There's been a little bit of basically attacks, but there's been reasoned rebuttal and the people who are actually posting projects, like Megazver, are clearly still able to do so; and I've issued a probation in the last few days for posting insults that went beyond critique of the technology or its uses and slipped into attacking posters. I would encourage you to try posting projects here too.
|
|
#
¿
Mar 27, 2023 20:54
|
|
|
Yeah I encourage folks to go there, but, not everyone is willing or interested in being restricted by D&D forums' formality/posting rules and I'm sensitive to that.
|
|
#
¿
Mar 27, 2023 22:56
|
|
|
A discussion of rules about posting about AIs has begun in the feedback thread, beginning with this post. I encourage folks to participate. Please try to remain civil.
|
|
#
¿
Jun 8, 2023 20:49
|
|
|
Well, I have to admit I'm a little surprised. I'm generally posting in favor of keeping this thread and challenging the arguments against use of AI art and text in the feedback thread, but if folks here feel like it's a lost argument anyway, I don't know what I'm supposed to do really.
|
|
#
¿
Jun 9, 2023 05:42
|
|
|
Several of you have posted in the feedback thread, and I appreciate that, thank you. This morning I replied to a handful of posts that I viewed as going beyond some acceptable level of "passionate debate" and just slinging insults - that's not OK in the feedback thread, and I'm not interested in creating a policy where that'd be generally accepted anywhere. To avoid repeating myself I'll continue to post my views and takes on the discussion, over there in the feedback thread. I'll just add that if you have something you want to say in the feedback thread but you don't want to post it under your account for fear of retaliation, you can PM a mod to paste it anonymously, as mentioned in the feedback thread's OP. If you prefer not to PM myself or Anti, some of the admins who pay closer attention to TG include sebmojo, Video Games, and hand knit, and you are welcome to use admins to provide anonymous feedback too.
|
|
#
¿
Jun 9, 2023 22:30
|
|
|
It's looking probable that we'll be rebooting this thread, and I'm looking for a volunteer to write a new OP for it. Please volunteer here or PM me if you're willing. The OP should introduce the subject (what are AIs, what are the current and upcoming tools, how do they work), and there will be rules written by the mods to be included in the OP. I think a fair introduction to both the tools and the ethical debate as it stands today is a good idea, because that will allow us to point to that summary and say "don't bother just repeating these points" to make it clear what the thread isn't for (re-hashing the same arguments already hashed). I'm not asking an OP to deliver final conclusions on each point of debate, but a reasonable statement of what the issues are and the fact that there is not a consensus on what to do about them. I'm thinking a new thread could go up sometime around Monday, but timing may change (and I'm still saying "probable") because Antivehicular and I have not yet reached a completely clear agreement, although we already have something of a framework and I think we're basically on the same page. We could hold off on a new thread for a bit if you guys want to workshop an OP for longer than that. Absent a volunteer, I can take a stab at a reboot OP myself, but I think some of you have a deeper understanding of the different tools so I think it'd be better to have someone who can describe the technical details.
|
|
#
¿
Jun 11, 2023 05:48
|
|
|
Raenir Salazar posted:If I'm reading Leperflesh correctly they're just asking for like a section to just mention that basically there is a debate, inform people of what the debate is, maybe list of points on either side, for the purpose of informing newcomers what the issues are and they should do their own research and reach their own conclusions rather than having that discussion in the thread. And as mentioned above the disclaimer section can also always point to the D&D thread. This is my intent, yes. Doctor Zero posted:That’s actually a good point. If the rule is you can get probed for off topic AI shitposting, you should also get probed for off-topic anti-AI shitposting. This should absolutely be the nature of the rule. My goal is not to settle a debate, it's to stop that debate from constantly making GBS threads up threads, including this one or its successor. I see this as not especially different from how I approach the fact we host threads for games made by unethical companies; I use moderation tools to make it clear that the people posting in the Magic thread do not have to answer for the crimes of Wizards of the Coast, but it is also actually OK to sometimes post in the Magic thread "holy poo poo they sent Pinkertons? gently caress these people" as long as that's not directed at specific posters. And yet, for a while a poster sported a redtext because they were perceived as having defended Pinkerton-sending, and that is kinda within bounds too, you know? (what they actually said was basically 'you should expect this, every corporation does this, why is anyone shocked' but their wording was open to misinterpretation, I think.) This is still SA, people gonna get mad and we do not have a kumbaya forum. So there's a gray area, mod discretion is in play, people may press the edges of it and I hope folks are prepared for some degree of rowdiness, as is the culture for SA at large. I cannot promise you folks an absolute safe space where nobody can ever say "boo." I can say that at least quite a few folks in the feedback thread expressed a desire to just not look in the AI thread and leave people there alone, and if that prevails, everything should work out, more or less. Megazver posted:Leperflesh, do you not see how this is getting more and more weird and discriminatory? And it's you who keeps proposing these bizarre new rule changes, like, what are you doing? I must admit, I am starting to feel less and less like "this must be so hard on you, you poor thing" at this point. I'm not, actually, the one proposing rules, but I have tried to restate the proposals that have been made. Most of it is rules for the whole forum, and the focus is on disallowing stupid recurring slapfights. But I would want to set something in place in the OP of a rebooted AI art thread that made it clear that there are strict limits on coming into that thread to attack its posters by repeating the already well-understood criticisms which, perhaps, we can summarize in the OP so that we can point to them and say "look this is already understood, please stop." And if what we try doesn't work, we'll try something else. Nothing is set in stone. The tools will evolve, the debate will evolve, people will evolve, we'll cross bridges as we come to them. quote:If someone wants to talk about how AI Art is Theft or Not Theft, this is also shitposting that's not allowed in the regular discussion threads. I understand that the people who are actually making GBS threads up all the discussion threads, and also for some strange reason you, don't seem to think that it's this poo poo that should be quarantined to its own thread. At this point, I don't see why I am supposed to care. If you can't moderate them, quarantine them to their own thread as well. Yeah I see no reason why that needs to be endlessly debated in Trad Games. There are other AI threads on SA, including ones in D&D, CC, and GBS, that seem to be chugging along fine and I think TG ought to be able to do the same. We have seen that the debate, even when closely watched by mods, descends into irreconcilable armed camps and absolutely nobody is being moved or convinced. Please believe me that my goal is not to have one side "win" here. If we need to establish a DMZ and the terms of a truce, then so be it. Leperflesh fucked around with this message at 00:11 on Jun 12, 2023 |
|
#
¿
Jun 11, 2023 21:05
|
|
|
What about something like:
Do you feel even this much language acknowledging that there is a debate is self-flagellation?
|
|
#
¿
Jun 12, 2023 00:10
|
|
|
Yeah a problem with Rutibex's rules is that he never checked with mods, he just posted them and then I guess expected mods to enforce them. Part of the reason I was suggesting a mod-written set of rules would be to make it clear this was going to be enforced. Raenir Salazar posted:I think the first point should be rephrased to be something like that, "by posting in this thread you're agreeing that this thread isn't about confronting people for utilizing generative AI tools." This sounds sort of like a EULA or something, I'd like to phrase this in a less legalistic sounding way. What if we just cut that first clause and say "This thread isn't about confronting people for utilizing generative AI tools". quote:I think the second point should be removed and changed into something like "Here's a list of tools that claim to be trained on content in the public domain, give them a try if that is something you'd be more comfortable with given the above mentioned concerns." Sure, and this is why I am hoping for a volunteer knowledgeable on the subject. The only one I know about is the Adobe tool. quote:"this thread is focused on the practical use of generative AI tools for roleplaying games." I'd also be fine with this, but I'd expand it to games in general - TG is about RPGs, boardgames, card games, and party games like Mafia, and I'd like this thread to serve all of these types of gamers.
|
|
#
¿
Jun 12, 2023 01:19
|
|
|
I don't love the word "containment" and "quarantine" isn't much better. IMO the best thing is to just have topical threads - eg the warhammer 40k thread isn't and shouldn't be the 40k containment thread, but if people constantly want to talk about fascism vs. parody in warhammer 40k and that was an issue around TG I might workshop a rule insisting that go in a specific place, too, and probably that place would not be in the chat thread, or the 40k thread, where people should get to post about their war dollies with some semblance of peace and take it as a given that we're not just harboring a bunch of fascists on the basis that they still collect space marines. Today, the target of ire is AI stuff, I'm sure there will be new things to be mad about in the future and if we develop a tolerable model for dealing with that, that'd be cool.
|
|
#
¿
Jun 12, 2023 06:38
|
|
|
I've rebooted the feedback thread and added new rules to the TG rules thread which Anti is just seeing, so she might want to modify them or something. If we've pretty much settled on language for rules for a rebooted AI for TG thread, is anyone willing to volunteer to be OP, do a little writup maybe on what the tools are, state of the art sorta thing? If not well an OP doesn't have to be super high effort but I'd love for it to be mostly about the thread topic with just a bit on the rules, rather than just be a post with rules.
|
|
#
¿
Jun 12, 2023 23:19
|
|
|
This might be a useful post to link to: https://forums.somethingawful.com/showthread.php?threadid=4000251&userid=0&perpage=40&pagenumber=311#post532264899 KwegiboHB PMed me and said they could help a possible OP out with technical details, they seem to know a lot about local LLMs so maybe reach out to that goon if you want!
|
|
#
¿
Jun 13, 2023 00:39
|
|
|
TG hosts both discussion threads (in TG proper) and game threads (in TGR). Typically you post a recruit thread in TGR, and list the game as recruiting in the recruitment thread, unless you already have players set up. Or you may just announce a game thread, like for a CYOA. Then interested people join up and you run your game in TGR. I think if you explicitly said in your recruitment thread that you'll be using AI assets in your game, people would either not join or not care, and then you could just run your game. But also I'm not sure why AI assets would be a highlight or feature, vs. the actual theme of the game and whatever is going on. It might also be fine to just not talk too much about what assets you're using, or poll players who have joined via PM to see if they object to the use of AI-generated character portraits or whatever. Ultimately I believe what triggered the big blowup and demand for rules, is people dropping AI stuff into discussion threads in TG, and that's what we've focused on. I don't recall anyone in the feedback thread raising the question of using AI stuff in a PBP or CYOA, which most people only watch if they're participating in, but that might just be because it hasn't come up yet. If there's a GM who wants to do that maybe they should just talk to a mod, we'd give it a try, and see if it's a problem or not. I can't imagine there's a big list of GMs just clamoring to do that.
|
|
#
¿
Jun 14, 2023 18:11
|
|
|
Raenir Salazar posted:So, I think this crosses over to something I was trying to point out, that usually for the purpose of transparency and ethics, people would like to be informed about these things. Hence why I suggested that people should be allowed to post if something is AI, because people have a presumed right to be informed, much like warning labels on food. I don't have a great answer for you because "source your art" isn't a hard rule, it's a suggestion, and it comes from people some of whom genuinely just think that's a good idea to do, and maybe some who presumably want to use it as a rule to expose anyone sneakily using AI art, and in the latter case I don't think that works or is practical and I don't think it's police-able either because "I made this without AI" "nuh-uh, you clearly used AI" is not a fight I want to try to parse especially as AI output gets better as most of us expect it to. I don't like some sort of "don't ask don't tell" policy either, seems bad I think it would be better to work from actual cases rather than try to theorycraft the cases like this. If you're not planning to run games yourself, then let's leave it till someone who actually is planning to run games and use AI art has an issue and mods can discuss directly with them.
|
|
#
¿
Jun 14, 2023 19:17
|
|
|
Humbug Scoolbus posted:That's a reason not to join a game? For some people it very clearly is, yes.
|
|
#
¿
Jun 14, 2023 20:16
|
|
|
I'm curious (coming from a cloud software background), is everyone running "local" models only on their home machines, or have people played around with running on cloud infra? Like there are cheap and even free infra options out there. https://cloud.google.com/free https://azure.microsoft.com/en-us/free/ https://aws.amazon.com/free/ https://docs.oracle.com/en-us/iaas/Content/FreeTier/freetier_topic-Always_Free_Resources.htm These are the four big guys as far as I know and they all have either a free trial or some low level of always-free compute resources available.
|
|
#
¿
Jun 15, 2023 17:31
|
|
|
Hey guys, I've been sick and basically checked out for a week, but I'm poking around now. How's a new OP going? I'd really love to get the stink of Rutibex off of this thread.
|
|
#
¿
Jun 27, 2023 17:37
|
|
|
|
| # ¿ May 22, 2024 09:27 |
|
|
Yup! Post it, we'll link to it from this thread & close the thread, bingo bango
|
|
#
¿
Jun 29, 2023 04:49
|
|