So fascinating to see such extremes of user feedback here.
My immediate wish is to see how you (and the opposing view) are prompting/writing to the AI. I can understand that one might adapt a certain comfort with one way of talking to a certain LLM and for that to disappear would feel very jarring. I've noticed enormous differences in different conversations and RP attempts with GPT vs. Claude. Right now, Claude Sonnet Extended (too poor for the $20/m upgrade) has been the best RP GM I've found, and I've been hacking away at this for years as a personal fun project.
The ability to create RPG system-specific skills in Sonnet on the fly and drop in "lore skills" and "GM skills" has been mind blowing to see actually work. I'm about 100 pages into a 13th Age session with Claude Sonnet and it's been really amazing to see how far this stuff is coming even at the free tiers.
I was going to post the same question just now, cool forum feature that you get pinged dupe posts... Followup question is, has anyone noted a difference with roleplaying styles? It strikes me as just another sheet of text the AI is supposed to care about but probably won't?
The AI just can't do this. It seems stupid, but LLMs can't "store" static data. You can continue reminding it of a piece of static data, but it can easily ignore it or concentrate too much upon some aspect of it for no reason. I imagine it would also start tapping stale data present in replies, i.e. it says "You have 55 gold." because user said something about having 55 gold somewhere on the rear edge of a context window, even though it was just reminded with subtext that user has 200 gold.
I like to point out to folks that pre-google Youtube was built on ffmpeg. You can set up streaming servers on it and all kinds of stuff. So who is stealing from whom... the pattern is theft from open source, not the other way around.
yes, I hear you. I have dreams of a publicly funded linux distro with accessibility and security (for elderly dudes and the like) the main design focus. My rebuilt federal government would develop this.
These features come from maintained modules or scripts that can be added if desired. There is a tiny bit of animation support in plain GIMP, mostly for working with animated gifs.
true that it wasn't good for mac. I gave up apple/mac and their increasingly shitty overpriced products 10 years ago. Since then Linux has come a long way and so has GIMP. Good enough to kill Photoshop? Not any time soon, but good enough for professional use certainly and good enough for new artists to start on. Install G'MIC and it's so much better.
some open source projects have very unpleasant communities around them, GIMP is not one of them. very easy to get into and everyone is extremely helpful and friendly.
Linux comes (at least on Mint and Ubuntu) with the poppler-utils suite of tools. you can merge, extract pages, even rip all the images out of a pdf in source quality. very useful.
Maybe you mean a more "brush and canvas" interface without complexity and distraction. I'm an artist that uses gimp. They are both great, Krita is just made with ease of use and emulation of irl tools in mind. GIMP can do emulation stuff too, but it can also do tons of other things, even video fx and animation.
I have been using "social software" for decades now. In the early 90s I ran a single line BBS networked across Canada and USA (every modem in the network called the nearest modem and we shared posts this way). I went to friendster (before that "makeoutclub", if anyone remembers that thing), myspace, then piled onto facebook like everyone else.
The only thing that seems to make a social media tool relevant is when a critical mass of young people, in that 15-30 demo, decides that it's time to adopt a new social media platform. This has been because they want a place away from their families, but there's also all kinds of memetic reasons behind it that have nothing to do with function.
Young people need to flood to open source, distributed networks like Lemmy. I don't mean to be a token old guy and speak poorly of an entire generation, but I'm not impressed with the embracing of Tiktok, Facebook and reddit. I find it hard to take anyone with a cause who inhabits these spaces seriously. I told my niece about why Facebook sucks, and she said "that's so boomer." Like, wtf?
So yeah. Young people, stop using twitter and all that junk. I promise if you come here I'll go somewhere else.
I believe you. have you tried something like making a fresh browser profile to ensure that it can't be looking at some stale txt file/cookie somewhere?
That is definitely disturbing, but I don't think the AI is remembering anything. Sadly, this AI will recycle the most weighted data from its training when the same stuff comes up.
I think the challenge is a balance between A and B:
A allows for much more creativity, but much less relevance.
B makes the AI seem more relevant, but much less original or creative.
If you were to push the model a little bit the other way, you might have a Text AI like the one here: https://perchance.org/ai-artgen
(click robot assistant button)
This AI will repeat things like normal, but it's like talking to a lunatic who just dosed 500 tabs of LSD.. It has no ability to focus on anything, & spews randomness from every corner of the universe without being asked to.
sounds like the "astronaut on a horse" problem; the ongoing challenge of getting AI to get a horse to ride an astronaut (it's always been easy the other way around). This might not be a problem anymore because AI now has lots of training data around this meme, I haven't tried in a year.
Basically there are not a whole lot of characters submerged in water for AI to train itself on. There's megatons of data of characters above water, by water, walking near water, floating on water, or swimming. It thinks you're talking about that data because it's the heaviest.
Firefox doesn't really need to make money so long as Google has to worry about being broken up by the Fed for antitrust reasons... Obviously so long as there is a national socialist/Trump regime in place that's not really something they have to worry about. Why, Goog even donated $5m to Trump's little Whitehouse evisceration/ballroom project, and will get their very own plaque in the ballroom that all the ballpeople can look at.
Google has donated over $USD 1B to Mozilla. The standard conspiracy fud is so that it can sink its claws into its open source competition, corrupting it with its omniscient googliness, but FIrefox barely blips on the radar of browser use, not even including Tor, Waterfox, Librewolf (which I'm currently using), and other forks. It's just not going to threaten any of the megas any time soon.
The actual motive for giving Mozilla so much money is simply to deflect any accusations that they're a monopoly that hurts market competition.
That's a weird one. The only times I've seen System in use was during chats with the Strict DM character in Ai Char Chat. It is used to run a (very broken) "PC inventory" script. I think it is supposed to have a specific purpose, it's not just an extra character.
If /nar doesn't cut it, you can always add another custom character to the chat. To my surprise they can actually retain some of their personalities when dropped into other character chats, but will need some more substantial lore written into their actual descriptions. They don't really retain anything like a history or /mem or /sum when used in other character sessions.
Oh, one thing I'm noticing, which actually touches on the racial bias issue, is that characters get "flushed" all the time. Even if they have dark complexions, the AI describes them as blushing all over the place. Sometimes they will "flush a darker shade of brown," which I'm not sure one can do (or can they?). And for no reason, of course, because it doesn't have reasoning or knowledge.
Another classic is that a character has far too many free hands, because the AI doesn't take into account what hands are, or what they're doing, or what they're for other than being contextually related to holding a thing and experiencing knuckle-related sensations.
Holy shit, yeah, I've gotten all of those in some variations. I actually have a hilarious session that was going sort of well for a while, using the Strict DM character in the default perchance AI Character chat. I wanted to play an Ian Bankes/Culture inspired sort of sci-fi adventure thing, just to test the tools.
AI kept telling me my scavenging party were entering "crystal caves," when nothing about crystal caves was anywhere on the agenda, and I began yelling at it and saying to stop talking about crystal caves.
It then doubled down and began saying things like, "The Crystal Caves sense your anger and begin trembling (or, as you say, "resonating") violently." I had to manually remove crystals from ever being mentioned in the sessions, in /mem and /sum just to get a crystal-free QA. Then it suddenly started telling me things like "the walls hear you scheming and begin to undulate violently," thus adopting the insane obsession with empathy, emotional terrain features; I made all the characters commit suicide. Any objection to me having them die was deleted and rewritten. It was the most meltdowny I got using this thing, way before I knew any of its limitations or anything.
Deleting anything the AI says more than once in a few pairs of replies helps a little bit. Bleaching /mems can sort of help too. I don't have any hope that AI that works this way can do anything more than what it's doing. It's not the way language works, but who knows. Ten years ago I was working on chatbots at megacorporations that couldn't even answer a single question with any consistent coherence.
Hey, thank you for giving the AI context window an actual byte size (if I am communicating what I mean in the right terms). I think I have a single chat that's exceeded 1mb in plaintext (maybe 2mb by now??) and it's sad to see it all fall apart like a person with a geriatric mental condition.
I have tried everything to salvage it, but cutting it up into bits and then giving the ai summaries of the stuff I deleted from its brain seems like one way to do it. That's kind of what you're saying, right?
However, there's also just something wrong with how the AI weights stuff, even in a fresh chat, the fact that I see the exact same turns of phrase in these posts, complaints share the same repetitions, points to maybe something wrong with the training data or something.
Has anyone noticed how characters tend to get obsessed with being proud about things? Or that everything starts sending shivers up everything's spines?
I forgot that I wrote this. Reading it again, seems kind of paranoid, but then...maybe it isn't.
So fascinating to see such extremes of user feedback here. My immediate wish is to see how you (and the opposing view) are prompting/writing to the AI. I can understand that one might adapt a certain comfort with one way of talking to a certain LLM and for that to disappear would feel very jarring. I've noticed enormous differences in different conversations and RP attempts with GPT vs. Claude. Right now, Claude Sonnet Extended (too poor for the $20/m upgrade) has been the best RP GM I've found, and I've been hacking away at this for years as a personal fun project.
The ability to create RPG system-specific skills in Sonnet on the fly and drop in "lore skills" and "GM skills" has been mind blowing to see actually work. I'm about 100 pages into a 13th Age session with Claude Sonnet and it's been really amazing to see how far this stuff is coming even at the free tiers.
I was going to post the same question just now, cool forum feature that you get pinged dupe posts... Followup question is, has anyone noted a difference with roleplaying styles? It strikes me as just another sheet of text the AI is supposed to care about but probably won't?
The AI just can't do this. It seems stupid, but LLMs can't "store" static data. You can continue reminding it of a piece of static data, but it can easily ignore it or concentrate too much upon some aspect of it for no reason. I imagine it would also start tapping stale data present in replies, i.e. it says "You have 55 gold." because user said something about having 55 gold somewhere on the rear edge of a context window, even though it was just reminded with subtext that user has 200 gold.
I like to point out to folks that pre-google Youtube was built on ffmpeg. You can set up streaming servers on it and all kinds of stuff. So who is stealing from whom... the pattern is theft from open source, not the other way around.
yes, I hear you. I have dreams of a publicly funded linux distro with accessibility and security (for elderly dudes and the like) the main design focus. My rebuilt federal government would develop this.
These features come from maintained modules or scripts that can be added if desired. There is a tiny bit of animation support in plain GIMP, mostly for working with animated gifs.
true that it wasn't good for mac. I gave up apple/mac and their increasingly shitty overpriced products 10 years ago. Since then Linux has come a long way and so has GIMP. Good enough to kill Photoshop? Not any time soon, but good enough for professional use certainly and good enough for new artists to start on. Install G'MIC and it's so much better.
some open source projects have very unpleasant communities around them, GIMP is not one of them. very easy to get into and everyone is extremely helpful and friendly.
Linux comes (at least on Mint and Ubuntu) with the
poppler-utilssuite of tools. you can merge, extract pages, even rip all the images out of a pdf in source quality. very useful.Maybe you mean a more "brush and canvas" interface without complexity and distraction. I'm an artist that uses gimp. They are both great, Krita is just made with ease of use and emulation of irl tools in mind. GIMP can do emulation stuff too, but it can also do tons of other things, even video fx and animation.
I have been using "social software" for decades now. In the early 90s I ran a single line BBS networked across Canada and USA (every modem in the network called the nearest modem and we shared posts this way). I went to friendster (before that "makeoutclub", if anyone remembers that thing), myspace, then piled onto facebook like everyone else.
The only thing that seems to make a social media tool relevant is when a critical mass of young people, in that 15-30 demo, decides that it's time to adopt a new social media platform. This has been because they want a place away from their families, but there's also all kinds of memetic reasons behind it that have nothing to do with function.
Young people need to flood to open source, distributed networks like Lemmy. I don't mean to be a token old guy and speak poorly of an entire generation, but I'm not impressed with the embracing of Tiktok, Facebook and reddit. I find it hard to take anyone with a cause who inhabits these spaces seriously. I told my niece about why Facebook sucks, and she said "that's so boomer." Like, wtf?
So yeah. Young people, stop using twitter and all that junk. I promise if you come here I'll go somewhere else.
I believe you. have you tried something like making a fresh browser profile to ensure that it can't be looking at some stale txt file/cookie somewhere?
That is definitely disturbing, but I don't think the AI is remembering anything. Sadly, this AI will recycle the most weighted data from its training when the same stuff comes up.
I think the challenge is a balance between A and B: A allows for much more creativity, but much less relevance. B makes the AI seem more relevant, but much less original or creative.
If you were to push the model a little bit the other way, you might have a Text AI like the one here: https://perchance.org/ai-artgen (click robot assistant button)
This AI will repeat things like normal, but it's like talking to a lunatic who just dosed 500 tabs of LSD.. It has no ability to focus on anything, & spews randomness from every corner of the universe without being asked to.
sounds like the "astronaut on a horse" problem; the ongoing challenge of getting AI to get a horse to ride an astronaut (it's always been easy the other way around). This might not be a problem anymore because AI now has lots of training data around this meme, I haven't tried in a year.
Basically there are not a whole lot of characters submerged in water for AI to train itself on. There's megatons of data of characters above water, by water, walking near water, floating on water, or swimming. It thinks you're talking about that data because it's the heaviest.
Firefox doesn't really need to make money so long as Google has to worry about being broken up by the Fed for antitrust reasons... Obviously so long as there is a national socialist/Trump regime in place that's not really something they have to worry about. Why, Goog even donated $5m to Trump's little Whitehouse evisceration/ballroom project, and will get their very own plaque in the ballroom that all the ballpeople can look at.
Google has donated over $USD 1B to Mozilla. The standard conspiracy fud is so that it can sink its claws into its open source competition, corrupting it with its omniscient googliness, but FIrefox barely blips on the radar of browser use, not even including Tor, Waterfox, Librewolf (which I'm currently using), and other forks. It's just not going to threaten any of the megas any time soon.
The actual motive for giving Mozilla so much money is simply to deflect any accusations that they're a monopoly that hurts market competition.
That's a weird one. The only times I've seen System in use was during chats with the Strict DM character in Ai Char Chat. It is used to run a (very broken) "PC inventory" script. I think it is supposed to have a specific purpose, it's not just an extra character.
If /nar doesn't cut it, you can always add another custom character to the chat. To my surprise they can actually retain some of their personalities when dropped into other character chats, but will need some more substantial lore written into their actual descriptions. They don't really retain anything like a history or /mem or /sum when used in other character sessions.
Oh, one thing I'm noticing, which actually touches on the racial bias issue, is that characters get "flushed" all the time. Even if they have dark complexions, the AI describes them as blushing all over the place. Sometimes they will "flush a darker shade of brown," which I'm not sure one can do (or can they?). And for no reason, of course, because it doesn't have reasoning or knowledge.
Another classic is that a character has far too many free hands, because the AI doesn't take into account what hands are, or what they're doing, or what they're for other than being contextually related to holding a thing and experiencing knuckle-related sensations.
Holy shit, yeah, I've gotten all of those in some variations. I actually have a hilarious session that was going sort of well for a while, using the Strict DM character in the default perchance AI Character chat. I wanted to play an Ian Bankes/Culture inspired sort of sci-fi adventure thing, just to test the tools.
AI kept telling me my scavenging party were entering "crystal caves," when nothing about crystal caves was anywhere on the agenda, and I began yelling at it and saying to stop talking about crystal caves.
It then doubled down and began saying things like, "The Crystal Caves sense your anger and begin trembling (or, as you say, "resonating") violently." I had to manually remove crystals from ever being mentioned in the sessions, in /mem and /sum just to get a crystal-free QA. Then it suddenly started telling me things like "the walls hear you scheming and begin to undulate violently," thus adopting the insane obsession with empathy, emotional terrain features; I made all the characters commit suicide. Any objection to me having them die was deleted and rewritten. It was the most meltdowny I got using this thing, way before I knew any of its limitations or anything.
Deleting anything the AI says more than once in a few pairs of replies helps a little bit. Bleaching /mems can sort of help too. I don't have any hope that AI that works this way can do anything more than what it's doing. It's not the way language works, but who knows. Ten years ago I was working on chatbots at megacorporations that couldn't even answer a single question with any consistent coherence.
Hey, thank you for giving the AI context window an actual byte size (if I am communicating what I mean in the right terms). I think I have a single chat that's exceeded 1mb in plaintext (maybe 2mb by now??) and it's sad to see it all fall apart like a person with a geriatric mental condition.
I have tried everything to salvage it, but cutting it up into bits and then giving the ai summaries of the stuff I deleted from its brain seems like one way to do it. That's kind of what you're saying, right?
However, there's also just something wrong with how the AI weights stuff, even in a fresh chat, the fact that I see the exact same turns of phrase in these posts, complaints share the same repetitions, points to maybe something wrong with the training data or something.
Has anyone noticed how characters tend to get obsessed with being proud about things? Or that everything starts sending shivers up everything's spines?