The absolute state of AI results.

[shovelquest]

check out this awesome looking game that i just discovered

[shovelquest]

god damn using chat GPT is fucking h ilariously horrible lol

it takes like 30 seconds for it to reply to every prommpt. I feel like its 2040 brutalist dystopia and im trying to get food out of the rusty food dispenser

lol we're so fucking cooked

[Ekco]

Originally Posted by shovelquest [You must be logged in to view images. Log in or Register.]

check out this awesome looking game that i just discovered

so many good metroidvania games i haven't played, like 30 of them.

i hate being hyperfixated on just a handful of games and never playing anything else.

[Ekco]

Originally Posted by shovelquest [You must be logged in to view images. Log in or Register.]

god damn using chat GPT is fucking h ilariously horrible lol

it takes like 30 seconds for it to reply to every prommpt. I feel like its 2040 brutalist dystopia and im trying to get food out of the rusty food dispenser

lol we're so fucking cooked

does ChatGPT not build a creepy profile on you, ive reset my Gemini one numerous times now because it starts to know me too well, the more i interact with it the better it gets algorithm wise

"write a quick summary of what you know about me"

Based on our past conversations, here's a quick summary of what I know about you.

You're a tech enthusiast who values open-source software and system customization. Your current setup runs on Arch Linux with the KDE Plasma 6 desktop environment. You're comfortable with the command line and prefer using yay for managing packages.

Your hardware includes an AMD Ryzen 5 9600X CPU and an NVIDIA GeForce RTX 3060 Lite Hash Rate GPU. You use this system for creative and recreational activities, specifically Stable Diffusion with ComfyUI. I have a list of the models you're currently working with, including lyriel_v16.safetensors and dreamshaper_8.safetensors.

In terms of personal interests, you've mentioned being into older MMORPGs and various genres of RPGs. You've also shared a Steam profile link, which indicates a preference for games like EverQuest and EVE Online.

In short, you're a knowledgeable and hands-on Arch Linux user who enjoys PC gaming and AI-driven creative pursuits.

and in my saved info i have this shit, so it knows not to bother giving me ubuntu or windows install instructions

[You must be logged in to view images. Log in or Register.]

so for the tech troubleshooting install vibe coding side of shit we're in drift mode usually and i don't have to scream at it, my last "remember this" cache was like hundreds of things

[Jimjam]

I used to use ai to summarise some findings of psychological, psychiatric, medical and social research papers. It wasn’t very good at it, because papers tend to say “previous researchers found x, concluding y, but didn’t consider z” but the ai would just tell you the current paper found x and y, ie say the opposite of what the paper was working towards as the ai didn’t report the paper mentioned x and y as a critique.

Despite that it worked kind of okay. Almost.

The problem is it doesn’t compartmentalise very well. It keeps a shadow copy of you and tries to force it even on to new conversations.

I’m not a programmer, but I did use AI to help me explore coding. I also had a Warhammer project on and used it to quickly reformat a table, i think an army list I drew up for a chaos dwarf project.

Now, due to lack of compartmentalisation if a conversation goes on for a few questions it tries to relate anything to Chaos Dwarfs or turn things into a program for mental health.

[Ekco]

Originally Posted by Jimjam [You must be logged in to view images. Log in or Register.]

Now, due to lack of compartmentalisation if a conversation goes on for a few questions it tries to relate anything to Chaos Dwarfs or turn things into a program for mental health.

yeah this is why i was exploring local ai, i assume the "memory" problem was just due to a scaling issue with millions of users so i got my local one writing everything to a PostgreSQL database in theory it can pull from for past conversations, i don't know enough coding and debugging something through natural language prompts is a mother fucker when it comes to the "is it pulling this information for the folder full of pdfs i supplied or just straight up hallucinating + whatever the model itself was trained on" so ive been trying to test on shit i actually know like DnD and EQ n shit with mixed results

[shovelquest]

Originally Posted by Ekco [You must be logged in to view images. Log in or Register.]

so many good metroidvania games i haven't played, like 30 of them.

i hate being hyperfixated on just a handful of games and never playing anything else.

I know we’re the first Roblox generation (everquesters)

[sammoHung]

Originally Posted by Jimjam [You must be logged in to view images. Log in or Register.]

I used to use ai to summarise some findings of psychological, psychiatric, medical and social research papers. It wasn’t very good at it, because papers tend to say “previous researchers found x, concluding y, but didn’t consider z” but the ai would just tell you the current paper found x and y, ie say the opposite of what the paper was working towards as the ai didn’t report the paper mentioned x and y as a critique.

Despite that it worked kind of okay. Almost.

The problem is it doesn’t compartmentalise very well. It keeps a shadow copy of you and tries to force it even on to new conversations.

I’m not a programmer, but I did use AI to help me explore coding. I also had a Warhammer project on and used it to quickly reformat a table, i think an army list I drew up for a chaos dwarf project.

Now, due to lack of compartmentalisation if a conversation goes on for a few questions it tries to relate anything to Chaos Dwarfs or turn things into a program for mental health.

Try Notebook.LM. You can upload as many papers as you need, and it will briefly and more importantly, accurately summarize them

[shovelquest]

I keep trying to complain about Ai in a post but t hen remember that Ai at its worst is still better than humans at their best.

[Ekco]

>Try Notebook.LM

that's what i use to turn Gemini's reports into ai podcasts hehe