pouët.net

Yeah, unfortunately. There are a lot of musicians and graphics guys "free" IMOE, but nobody asks them..

tht was not _exactly_ my point here

Lots of hypotheticals, lots of "you have been sleeping, time to wake up, AI can already do THIS" examples, yet no example of anyone using LLMS to actually make something for real. As with every discussion around LLMs for programming I've seen, heard or partaken in over the past couple of years. When you press someone for a practical real world example of something they've made using an LLM it always turns out to be the most mundane, unimpressive shit you can imagine, but for some reason the narrative is always that using LLMs for programming somehow "changes everything".

I'll consider maybe believing at least some of the hype when I see it actually change anything.

people who like AI and LLMs are spending too much time on twitter replying "interesting…" and "this changes everything" whenever someone just post a new "open source library"

Oh, please...can you at least use AI assistance for a while to formulate an opinion or just don't share your opinion? People heavily criticize AI tools, but don't even try them out.

If this is "how we roll", of course there will be huge discrepancy in opinions.

Oblivious

Ha.

It will not change everything, but it will change something. It already did. Ha!

I tried make code for demo with AI.

Results are horrible... Don't try this at home!!!

ChatGPT is coder... Sorry, no video yet

You don't need to smoke to know that it causes cancer.

I for one would like to see an AI powered prod with Blueberry and/or Azure at the helm!

the point is: you won't. most parties nowadays have a strict non AI rule. that is the point of the thread. blueberry claims that the current ruleset was established with mostly graphics and music in mind (which I think is correct) and that we should have a discussion about AI usage in code with regards to that ruleset (which many but not all parties use).

even now we can see that this ruleset does not fit. e.g. blueberrys entry at evoke was allowed even though some tooling had been developed with some AI assistance ((I assume from his comments that this means that during the development of the no-cpu framework he had AI autocomplete on).

my latest entry at GERP also had some AI usage (which was properly disclosed). I recorded myself with a webcam and then used google mediapipe to extract my face. I would say that this would count as AI usage - I could have instead manually created something in blender or so. GERP also had a strict no-AI rule and I was fully prepared to be disqualified. However when showing the orgas the entry there was not even a discussion about it as it seemed ok right away. btw. during development of the preprocessing toolchain I also had Copilot autocomplete on but judging from blueberrys experience at evoke I thought that seems ok (and none of this code made it into the demo itself). So far copper showdown editor does not have any AI assistance :)

So it seems the "no AI" ruleset does not even reflect what orgas actually meant. probably there are more examples.

I’ve noticed one thing about AI. I’ve been trying to create a demo using AI tools, but I keep running into the exact same problem. When I ask the AI to generate 10 fractal images (just as an example), it maps them completely wrong, with incorrect colors. The result has no real aesthetic value. It seems to me that without human intervention, it can’t produce a high-quality result.

But you do need to smoke to discuss cigarette brands and flavours.

Thanks for that analogy. It's very fitting imho. See there's no healthy cigarette brand or flavour afaik. If you want to discuss cigarette brands and flavours, sure give yourself cancer. If you don't like cancer, steer clear of cigarettes. Pretty simple. Now just replace "cigarette flavours" with "LLM art&code", and "cancer" with "brainrot". You're sorta genius man.

dude's like "try LLMs before having any opinion."
but also, studies say the more people use LLMs the more they become relient on it, and it increases overconfidence.
+ as a friend from the scene told me yesterday: you cannot change one's opinion with studies.

So I'll just stick to my gun:

Excellent analogy! So everyone who wants to stay clear of cigarettes, stay clear of cigarettes. Now just replace "cigarettes" with "LLM art&code", and "cancer" with "brainrot" :-)

100% agree here.

Yeah, people became soft indeed. Omg, AI will give me brain rot, I'm so afraid.

you talk like you've achieved a lot of things already; you seem to think you're already at the top of your game. I wake up everyday thinking I havent peaked at anything, I look at your productions and think: this is mediocre, very very mediocre; but you're talking like you know your shit and you are superior, I envy your confidence.

Also, people who say that other "people became soft" are rarely the kind of best people.

Remember that this guy was rooting for Skynet in Terminator movies and he sincerely thought that we all were. I shit you not. It's all there in writing.

Everytime I go on pouet…

I do agree that we should regulate AI the same way we do with cigarettes - extensive publicly funded studies about the dangers, warning labels highlighting the severe consequences, banned for underage consumption and especially from public places, normalized social stigma if you use it among non-users...

You see how well this analogy works. Cigarettes were just banned at the Deadline party. If you want to smoke, you need to go outside.