|
I made an infinite How It's Made ripoff with a questionable image diffusion API: https://www.youtube.com/watch?v=xu6VJhFHueE https://www.youtube.com/watch?v=ZbpeedkuEnM https://www.youtube.com/watch?v=zQJtY7EhmaA It would be cool to have procedurally generated music too but I don't know how to do that.
|
# ¿ Feb 15, 2023 12:01 |
|
|
# ¿ May 14, 2024 00:48 |
|
Seeing AI become a real thing is so surreal. It has been this far-off sci-fi concept for my entire life, and then over the course of a couple years suddenly it starts to become concrete and real (while also making the future so much more uncertain). It seems like for the foreseeable future there will need to be humans in the loop with both AI art and AI text unless the goal is bizarre hallucinated AI nonsense. The "foreseeable future" keeps getting shorter though. On the topic of bizarre hallucinated AI nonsense, I made an episode of How Is It Manifested?? about Evil Clowns https://www.youtube.com/watch?v=ME9AvjPfYuY and another one about Demonic Relics https://www.youtube.com/watch?v=JHzUBbP381s They generate faster than the time it takes to watch them, so in principle it could run forever. Fun stuff.
|
# ¿ Feb 15, 2023 16:11 |
|
pixaal posted:Make a Twitch channel and profit from the ads. Strike while the iron is hot, get them a bit more interesting every run and let it run for the internet at large. I'm not sure if there is a good way to do that right now without playing it on a delay with someone moderating them in real time. The Seinfeld thing showed what these things can spit out on their own, and that wasn't even using image generation. I don't really give a poo poo if people see some weird AI boobies or something but Twitch has shown that they have no patience for AI stuff running wild and it seems like if you can't do it on twitch then you might as well not bother.
|
# ¿ Feb 15, 2023 16:34 |
|
lunar detritus posted:Eh, I'm still not convinced that the Seinfield thing wasn't on purpose. A basic word blacklist could solve most issues, with a normal prompt AI is not going to do transphobia or racism via metaphor so you can just blacklist controversial words and slurs. And you can add an automatic nsfw filter to the images. It's not completely fool proof but I think it can be made safe enough for twitch. pixaal posted:Stable Diffusion's Not Work Safe filter is super aggressive too, it loves to classify Shrek as a dildo. If you turn that on you are probably pretty good. That ban is only 14 days that's really not the end of the world that's a good chance to review your systems and help make sure it doesn't happen again. I'm sure it will escalate if they start loving around. Hmm yeah good points. I was using some random txt2img api but I switched to the official stable diffusion one and it seems like they have more protections to prevent bad poo poo showing up. I'm using GPT for the dialog so I could just run that through the GPT moderation layer and I'd be confident enough about that. Seems like the image generation would be the biggest cost, at around $10-20 a day if it ran 24/7 with like 6-10 images a minute.
|
# ¿ Feb 15, 2023 18:02 |
|
How Is It Manifested is now streaming: https://twitch.tv/HowIsItmanifested It's only doing reruns right now because I don't want to burn through all of my credits *right* away, but there are like an hour of them in there from some testing yesterday. No idea how long it'll run, this is the first stress test. The episodes it has right now are a mix of "random" ones and ones that were added manually to test it.
|
# ¿ Feb 16, 2023 18:24 |
|
Doctor Zero posted:Doctor Zero is taken by a Strange Mood. This is probably the best AI thing I've seen. I've been adding a bunch of poo poo to my AI show. Just hooked up google text to speech stuff, which is a huge improvement (if a bit of a hassle to set up). https://twitch.tv/howisitmanifested
|
# ¿ Feb 18, 2023 17:13 |
|
Someone used the command !chron World War 2 Air Dogs and a minute later we watched this: https://clips.twitch.tv/AnimatedTardyNightingaleWoofer-cn_K4Z0xm2t3NPh6
|
# ¿ Feb 20, 2023 14:42 |
|
LASER BEAM DREAM posted:I had ChatGPT tell me a little more. RIP Chips Went out as a good boy.
|
# ¿ Feb 20, 2023 19:04 |
|
Stable Diffusion has a cheap API available, which is a point in the column for that.
|
# ¿ Feb 22, 2023 01:50 |
|
https://clips.twitch.tv/SpikyMotionlessTurnipUWot-b-HWCLMoAk480xr_
|
# ¿ Feb 23, 2023 17:48 |
|
I'm so happy for them.
|
# ¿ Feb 23, 2023 22:37 |
|
"Baloo holds a wrench, standing in front of a rickety old machine" Weird al Ratatouilled "grimace from mcdonalds, collodion photography" Beetlejuice Dark furby "The Joker, Gilbert Gottfried, holding a bag of cash, standing in front of a burning building, collodion photograph"
|
# ¿ Feb 24, 2023 15:39 |
|
Soulhunter posted:I've been playing around in Midjourney with 'muppets' as a keyword. The muppets can make anything kid-friendly, right? Right? Oh my goood. AI makes such good muppets. I want to watch every one of those movies. The Schindler's List one in particular is absolutely inspired Lucid Dream fucked around with this message at 19:52 on Feb 27, 2023 |
# ¿ Feb 27, 2023 19:50 |
|
IShallRiseAgain posted:I actually was already working on that These look great. Definitely check out the elevenlabs pricing, because I was surprised at how quickly it becomes prohibitively expensive. Definitely too expensive for my stream, at the rate I send TTS through it. It wouldn’t be too bad as a last pass on an otherwise good video though, so if you do rough cuts with something else and then upgrade the audio on good ones then it might not be super expensive. pixaal posted:dwayne the rock johnson muppet AI muppets are the best.
|
# ¿ Feb 28, 2023 14:37 |
|
https://clips.twitch.tv/FlaccidUninterestedChimpanzeeStinkyCheese-_WTsJogCVQupRVUF
|
# ¿ Feb 28, 2023 20:06 |
|
IShallRiseAgain posted:So, I'm messing around with animation stuff. Really awesome stuff. I love the zoom with separate foreground and background layers, very cool.
|
# ¿ Mar 5, 2023 18:00 |
|
Dialog for games is tricky because you run into an interesting problem where you have to constrain the output so much and feed it so much context that it’s not even clear how it would save any time or effort compared to just writing it all out by hand. Let’s say you have a fantasy game, and a player asks about the king… if you don’t give the NPC context information about the king, then the AI will hallucinate info, and even if you fed the AI a Wikipedia page about the king, users would be able to probe the edges of the information easily with their own questions. You could (in principle) have the AI just not answer if it doesn’t have sufficient context… but now you’re constrained completely by how much context you can feed into the NPC. Maybe it will work some day when these models can fully internalize a huge wiki full of characters, events and connections.
|
# ¿ Mar 6, 2023 16:37 |
|
pixaal posted:Depends on what kind of story / game you want to make. If it's open world it doesn't matter if in my play the kind turned out to be a dragon just because I spread a rumor that he was around the world and in most games he's a demon or something. You need models for all of these unless you are having an AI make them on the fly. In general the problem isn’t so much about what information the NPC spits out as much as the consistency between conversations and NPCs. Like, real time dialog with an NPC isn’t very interesting if the things they say are inconsistent with the state of the world or other NPCs. Solving the memory/context problem is one of those Hard problems that will be solved eventually, but existing tech won’t cut it I think.
|
# ¿ Mar 6, 2023 16:50 |
|
Deki posted:Depends on if you want the ability for the player to ask entirely freeform questions or not. You could seed the context with points of interest and stuff, but in your example it works fine if the player asks that specific question, but what about like… “has Dracula drained any of your loved ones?” There is no good response for that unless the NPC has relevant context. They can’t really respond “I don’t know” and if they answer anything else without sufficient pre-fed context then it is going to just make poo poo up that might be inconsistent with other established facts/events in the game. GPT is pretty good at maintaining context within a conversation, but using that in a game with actual constraints and static story/character/etc elements is where it falls apart (Right now). Lucid Dream fucked around with this message at 17:07 on Mar 6, 2023 |
# ¿ Mar 6, 2023 17:03 |
|
pixaal posted:Well if you are running it local you just have the AI check the game state, and if it's unknown it's being decided on right now. Doing it purely in GPT would be silly. The AI “checking the game state” is really the problem though. I don’t think NPCs with real time dialog will work well until there is a way to build all of the relevant context for that NPC dynamically based on the state of the world. It’s the kind of problem that is solvable in principle, but also the kind of thing that very smart people in research labs are working on (and haven’t solved yet).
|
# ¿ Mar 6, 2023 17:11 |
|
pixaal posted:Talking about building a large scale game with current tech is silly you have no room for anything but the AI and will require very powerful GPUs the cost of production and how niche a product this will be with most people going "yeah but it looks like poo poo and I need a 4090", this is 5-10 years out at a min, all of these features are things that have to be solved before it's viable and they can be. The problem right now is that “check the state” means something very fuzzy, and turning that into a text prompt in such a way that actually provides sufficient context for a conversation is a Hard problem. Internal consistency within one chat isn’t really hard, but it’s the consistency with the rest of the world that is hard. With existing systems, you either have to build with the assumption that the AI will hallucinate stuff (not an impossible gamedev challenge, but one that implies a lot about the kind of game it can be) or you have to constrain it so much that you might as well just have multiple choice options rather than real time dialog. Again, all solvable problems, but what I’m really getting at is that real time dialog *looks* much more promising than it really is *right now*. AI is great at hallucinating things, games are great with concrete things… that translation from one domain to the other is where the problems lie at the moment. Lucid Dream fucked around with this message at 17:32 on Mar 6, 2023 |
# ¿ Mar 6, 2023 17:29 |
|
JazzFlight posted:Kind of related to the video game discussion, how do you think they automate the endless Steamed Hams stream? Outline plus either asking GPT for random stuff or from a pool of things. You could ask GPT to come up with a random food or feed it with a food item from a big list. Given that the structure is always pretty much the same, I’m guessing that they have a prompt for each “scene” and they string them together to make an “episode”. Edit: I assume they’re using a pre-made outline, because it would diverge a lot more if they were actually generating the outline itself on the fly. Lucid Dream fucked around with this message at 18:30 on Mar 6, 2023 |
# ¿ Mar 6, 2023 18:26 |
|
feedmyleg posted:When it comes to writing something coherent to a consistent formula, I've had success in having it go from more broad to more granular. I ask it to write a logline, then summary in a 3 act structure, then each act in more detail. I could imagine one of those infinite sitcoms doing much the same. Yeah, you definitely can start from a simple prompt and inflate it into a whole episode by moving in steps like that, but since steamed hams is using pre-made assets they can’t actually diverge from the formula in any meaningful way. I really want to try making something more long-form and open ended and tie it into stable diffusion and see what kind of nonsense comes out the other end.
|
# ¿ Mar 6, 2023 18:42 |
|
Doctor Zero posted:I don't think that the problem is all that hard, and we'll see some great applications of the tech within a year or two. Think of how fast ML is advancing. It's crazy. I've never seen anything like it. The Internet was close, but poo poo, this is going faster. Don't get me wrong, everything in your post is doable eventually, I'm just noting that while the current chatbot stuff *seems* really good for games, there are still some very hard problems to solve before they can realistically be used for real-time dialog. Right now it is sort of a mirage, where with very little effort you can make something that *seems* amazing, but the more you work with it the more you realize how unsuitable it would be out of the box. Edit: Worth keeping in mind that figuring out how to do long term memory for these things is like... one of the major steps between where we are now and real AGI. Lucid Dream fucked around with this message at 21:03 on Mar 6, 2023 |
# ¿ Mar 6, 2023 20:59 |
|
Doctor Zero posted:What I meant was that you might not need to have long term memory. From what I understand about GPT-3, each session you start with it is distinct with no history of anything else. In playing with it, it has a better memory than AI Dungeon, but yes, it’s still limited. Keeping track of events that pertain to the NPC would mean you could have each instance ‘load up’ the history and then go from there, kind of like telling ChatGPT ‘you are a homeless one armed Indonesian on Mars as a miner of mars rocks. Go!’ The issue you'd run into is a limited working memory. ChatGPT kinda handwaves it away with some stuff behind the scenes, but the model can only handle a total of 4000 tokens, which comes out to around 3000 words. That includes the input and the output. So you can't simply stuff all the chatlogs in there, you'd have to summarize and extract key events or something. You also then have to feed it all the *other* context about the world, as well as any other instructions for the model to use. Rumors are that GPT-4 will have up to 32k tokens, which would reduce some of the bottleneck, but it doesn't really fix the underlying problem. Not to mention, more tokens in the input + output = higher cost, so a big giant wall of context for the AI so that it can respond "beans" to my latest question of "what is your favorite food" would cost a non-trivial amount of money.
|
# ¿ Mar 7, 2023 00:30 |
|
Telsa Cola posted:I realize it's difficult to predict, and kinda dumb question but where does everyone see all of this stuff going in 6 months to a year. It seems to me (who checks in and out with this kind of stuff) that since a bunch of stuff got leaked or publicly released shits been moving in leaps and bounds. I'm not sure on timelines, but I think we're going to see two diverging paths with AI stuff, where one path is really good text generation, really good art generation, etc but each of those things happens in a vacuum so they don't really talk to each other perfectly. Another path is multi-modal models that can use visual/audio/text as input and output. If GPT4 can do 32k tokens then it might be able to output an entire screenplay. If it's also significantly smarter than GPT3, it might even be a *good* screenplay. Some day the multi-modal models will be able to generate a screenplay and then make it into a film, but that's not going to happen in the next year.
|
# ¿ Mar 7, 2023 01:14 |
|
Ovenmaster posted:What makes me excited about these new AI tools is that they could potentially help bridge the gap between the procedural and the tailored. Even if the initial implementation will be crude, it would be a big step up from the soulless empty terrain with copy/paste towns. A few more years down the line, who knows? We might even have a Skyrim the size of Daggerfall where every NPC has voiced lines and offers multi-tiered immersive quest-lines. It's still hard to imagine, but with the pace things are happening I wouldn't bet against it. I think that all of these AI systems shine when you use them to rough out a first pass of something that can be refined. It's more obvious with the art stuff, but it really applies to all of it. It's not hard to imagine a workflow for an open world RPG that involves prompting an AI system with high level concepts which are inflated into a first pass of some content. You might prompt the system with "I want that cave to be a home for a family of kobolds who stole a relic from a nearby necromancer." and it could inflate that into NPCs, loot, combat encounters, dialog, etc. It might generate a stupid relic, or it might generate an NPC with dialog that isn't internally consistent with the rest of the established world... but as long as you can correct those issues easily you're still saving an incredible amount of time and effort. Humans can focus on the high level stuff, and computers can do all the busy work.
|
# ¿ Mar 7, 2023 09:02 |
|
LASER BEAM DREAM posted:My attempts at getting LLaMA to act like a crude ChatGPT are going great... Perfect, no notes.
|
# ¿ Mar 8, 2023 19:12 |
|
IShallRiseAgain posted:so I've been messing around with prompts, and I think you can use it to make more guided text adventure games. This is pretty cool, are you doing it entirely through ChatGPT? I've been working on something sorta similar with the ChatGPT API but more focused around building more traditionally structured game content with the generative text and it's going pretty well. Right now I'm working on generating random oregon-trail type events for any arbitrary location, and it's such an interesting design challenge. I had to first generate the locations, then determine stats for them, and now I'm working on building a list of hazards (falling rocks, etc) that will be turned into random events in locations that are considered hazardous. It's such a wild design challenge, because you have to somehow wrap your head around turning infinite procedural options into something concrete to use it for traditional videogame mechanics. There is this real diamond shaped thing to it, where you can start with a simple prompt, inflate it out into a world with points of interest, and then crunch that generated point of interest down into concrete statistics that the game can use for normal gameplay purposes. Edit: There is a real interesting tradeoff though. Like the structured approach lends itself well to integrating traditional gameplay mechanics, but you're also crunching the actual interaction back down to normal videogame levels. A game entirely within ChatGPT has that benefit of having a lot more room for emergent gameplay. Lucid Dream fucked around with this message at 21:39 on Mar 30, 2023 |
# ¿ Mar 30, 2023 21:25 |
|
IShallRiseAgain posted:Here are the very early stages of my AI powered text adventure game that is more guided. https://www.youtube.com/watch?v=uohn5o0Cgpw. It looks like its going to work. its going to involve a lot of effort, but I think it will pay off in the end. Hey this is looking really cool! I'm super interested to see where you go with this. I'm really glad someone else is working on AI gamedev stuff, it's started feeling really lonely considering how much potential there seems to be in it. Lucid Dream fucked around with this message at 21:44 on Apr 2, 2023 |
# ¿ Apr 2, 2023 21:42 |
|
RPATDO_LAMD posted:because none of the limitations or safeguards are actual limitations or actual programming, they're just more text fed in at the start of the text-generation model's memory and it treats them the same as any other text it's asked to generate completions for The ChatGPT/GPT4 APIs have a way of breaking up the prompts so that you can specify a "system" message that has a lot more weight. They also let you mark certain parts of the prompts as from the "user", and all of that stuff is intended to be used to train it to be harder to do prompt injection. At this point it's still unclear as to how effective it'll be in the long run.
|
# ¿ Apr 3, 2023 03:55 |
|
It is a sad reflection on the human species that it is a problem for AI to replace human labor. The silver lining is that AI doesn't discriminate, and it's coming for everything. This isn't going to be something that people can ignore or minimize, and it's not something that affects a specific socio-economic group, race or religion. We're all in the same boat, and we're all going to have the same concerns. We're not going to see literally every job replaced in the immediate short term, but everyone is going to be affected before long. Given recent history, I doubt everyone will agree on the specific implementations of what comes next, but we will see movement.
|
# ¿ Apr 7, 2023 18:46 |
|
Analytic Engine posted:Distinguished Goons and Goonettes, You've got my interest piqued, sounds neat. I'll be there. Lucid Dream fucked around with this message at 10:27 on Apr 8, 2023 |
# ¿ Apr 8, 2023 10:18 |
|
Suno is pretty great: "Eggs" https://files.catbox.moe/3ftw5z.mp4 "Goblins" https://files.catbox.moe/fvueg4.mp4 "Poorly Drawn Horses" https://files.catbox.moe/lzhi98.mp4
|
# ¿ Nov 25, 2023 03:45 |
|
"Heed This Broadcast" https://files.catbox.moe/potyfj.mp4
|
# ¿ Nov 25, 2023 05:27 |
|
https://cdn1.suno.ai/ec774e8c-7f0a-49a3-808e-fa974791d178_1.mp4
|
# ¿ Nov 26, 2023 18:41 |
|
|
# ¿ May 14, 2024 00:48 |
|
I made a generative AI SCP thing https://www.theanomalypapers.com/ it might be up and down as I fiddle with it, but it seems to be working pretty well so far.
|
# ¿ Mar 21, 2024 01:06 |