enthusiasm_headquarters @ enthusiasm_headquarters @lemmy.world

I just want AI to be my buddy

---

Read more

Posts

3

Comments

80

Joined

12 mo. ago

Yes, I've definitely figured out all your very good points on my own. Still, such subtleties such as "-" introducing a negating effect or a missed piece of punctuation being a big deal is good to be reminded of. Have you noticed any difference with commas? I notice I have some phantom commas, i.e.: "she's wearing black, ,holding a dagger". I wonder if these are changing the renders, I just assume its being interpreted as noise.

"It's janky. It's charming. It works. It's Copyparty." ®

I worked for an early AI project for [large tech giant]. While there, Alexis Ohanian came into the office and talked to my boss about using reddit's content (your posts and probably private correspondence, too, we got it from every other social network) as corpora for the AI/bot, it wasn't quite "training data" yet, it was early still in the game, but most likely it has become such.

So yeah, add that to reasons to hate reddit. I have never understood the appeal of what is essentially a for-profit USENET/BBS system, except that you can decorate yourself with flare or whatever it's called. Why'd people start using it? Why do we give away our collectively-owned technology to these dipshits?

Um... haven't you seen the Demon Seed?

Are you trying to get us all killed? https://www.youtube.com/watch?v=H6O1NRs-YuU

"Barack Obama committed treason... That's why I hired a bunch of prostitutes to talk shit about him while they pissed on a bed he slept in as Russia taped me.." - Donald Trump.

assume AI is stupid,

Definitely don't have to twist my arm on that one.

Using a dash (-) between words worked for SD, but Flux can be quite quirky (using the ‘-’ as a negatory?).

Holy crow, this is really friggin useful to know and I probably would never have figured it out myself. Thank you! I wonder how many renders I've ruined by negging things I actually wanted to emphasize...

what sort of machine should you have to try this out? I have an older PC running Linux, unfortunately.

I'm definitely not even close to the last word. maybe there's a script out there that extends the features? I do not know.

Hey, very good catch with the 'old' thing. I've noticed it can hone in on such abstract details and emphasize them. Sometimes I have to avoid using the word girl or young because the AI will go a bit too "girl" and wayyy too "young." Like a 30-year old woman with a 3-year old's face kind of thing.

I've not tried editing the summaries, but have edited the much larger /mem output as well as adding lore to the lore text files, mostly just to see if I can stop characters from using the same words over and over again or falling in love with some meaningless detail that only they care about.

I honestly don't know if it helps. The way "memories" work with textual AI makes no sense when compared to actual memory. Even if the lore and memory is spot on, the AI won't understand how to contextualize a passage (or when to consult it in the first place) with any coherence. Despite a concept being repeated throughout the lore and mem files, I still have to remind the AI characters about things.

I've also used the Reminder feature with mixed results-- mostly not worth using. At times the AI will start reciting the reminder. If you put ([AI]: Reminder: Don't say "But tell me," or "But hey," anymore.) in its reminder field, it might immediately say something like:

"AI Guy makes sure not to say 'But tell me' or 'But hey,' knowing how much this irritates the MC. 'But inform >me, what do you like to do for fun?' AI Guy asks.

... or something annoying like that.

I've also never actually seen my lore file being used even after chats reach 2,000 lines, but maybe I'm missing something.

I'm not saying these things do nothing, I just haven't noticed what these features actually do. I think it's just the limitations of the current model. If you are able to do any of the above and see results, please share.

I don't think there's a default solution to this. The org abilities are pretty sparse rn. You can make folders. move chars to them, search for text across conversations, import and export. That's pretty much it.

Because maintaining the USPS is a federal obligation as stated in the fucking Constitution. Anyone that doesn't understand this hates America and has AIDS.

There's absolutely nothing wrong with Mint.

There's a small army of Linux "snobs" that look down on it for recondite and mostly silly reasons. Mint is a great and user-friendly OS. The only thing I can say against it is that many of the binaries in the distro app manager are very out of date, but this hardly matters now because AppImage and Flatpaks are so on top of it and great.

without any examples of errors or what you're trying it's very difficult to help.

Unless I'm mistaken, the OP wanted a diversity along the lines of the same prompts, not simply randomized output. All of your images there are tapping into vastly different domains. Try describing a human being with the constraints of a scene, clothes, a particular action, and that's where it can get homogeneous.

Yes, I have had some success with it. It's rather difficult to simply reproduce an issue like this. If you have a prompt where you are getting a lot of the same faces, maybe you can post it here and I can try what I've suggested on it or see if I can help.

Off the bat, the Professional photo preset has a lot of same-ish output. Using 90s photo or Casual Photo, I assume because the training data (social media photos, family, school & work snaps etc.) is just so incredibly diverse, will by default tend to offer more variation.

it's sad, pathetic, and stupid that one has to download a potentially dangerous hack to do something so basic.

yo compiz is the shit. You can do ANYTHING with it. It took me a while to figure out because where the hell is the manual, but I have my own custom thing going on and it's brilliant.

I think the fastest way for linux to spread are a) a state-sponsored (totally open source) product that sees a free and open OS as part of a commitment to a free and open society. or 2) one of these fuckhead billionaires drops $200M or so into a trust, rather like the Poetry Foundation, which has the singular commitment to create an OS for people and to support it indefinitely.

I don't think the answer to any of society's ills is to get Wallmart involved. ed: walmart however its spelled WGAS.

congratulations, you now walk with the righteous.