Wow, I didn't expect so many artists to respond to a programming thread. Super cool. More of this!

With that in mind, maybe I'll throw a different question into the mix for artist with no coding background actually reading through the thread:

Have you tried using chat models to help you with the command line? To write tiny one-off scripts? In general, are they less annoying to deal with than asking one of us ********* about what flags to use with ffmpeg?

Nope, I never used AI, except for when it was just getting started. I have an intense hatred for AI, and promised to myself that I will never use it. Beside, it's associated with the people that I abhor the most in this planet.

I'm experimenting with using chatgpt as a game chat moderator, and I'm also toying with the idea of using it to simulate characters, but the prompting is trickier.

I sometimes ask AI to help me with my game dev math

I think the reason why art should be defended against AI as opposed to literally any other industry is that art has been the primary form of human expression since the dawn of time. It’s time-consuming and requires incredible skill at the craft, but learning to express yourself through these skills is what makes art so powerful. When AI generates a whole song off of words though, there’s nothing human about it. It loses all of its expression, and therefore we lose the point of having art in the first place. Plus, it invalidates all of the work artists put into their craft.

I can’t really explain it either. Plus, I absolutely have a bias. But generative AI art has always felt a bit different to me than AI agents, or generative AI writing code. It’s more like an efficient tool in those cases than an emotionally obsolete replacement.

Weird Al Yankovic ain't never replacing me

Interesting. I've also been thinking about using it as a moderation aid (sentiment analysis). Are you using a local model or an API and what libraries if any?

Where's the fun in that?

The server makes api calls to gpt 4o nano, it reads the message and if the message is problematic, recommends either a warn, mute, or ban.

It looks like you're referring to a discussion or article about replacing oneself with AI. While I couldn't access the exact content you mentioned, I did find a related discussion on Reddit where someone shared their experience of automating parts of their work using AI. There's also an article exploring how AI can replicate personalities to create digital clones.

Are you thinking about using AI to automate aspects of your work or daily life? It’s a fascinating topic!

Very limited use cases right now. Not reliable for research, not reliable for nuanced and involved programming, not reliable for creating art without putting in legwork yourself (not that you should be using it to “create” artwork, animation or music without any effort)

I had a subject at work about that, I was ask to research how an AI could help me in my work. Since my company is very selective on confidencial security, they have implemented an Llm (llama) that can only work within the intranet.

Turns out, not so much.

• Generated code doesn't respect environnement restriction, can use deprecated method (it's learning is only from end 2023) and produce a ton of little mistake/bug
• It can help showing you method to guide, but I know well my library (restrictive environnement lead to little number of dependancies), so I don't need help with that.
• Helping me understand other's code is where I use it the most, specialy with weird programming langage, but with récent event, it showed me that its limits comes quickly...

Tldr: not so much

why? if you are programmer, ostensibly going to land a job in modern tech- you're not doing yourself any favors. Like being too stubborn to go from a text editor to becoming familiar with a proper IDE, or insisting on using assembly to write a Hello World program when there's higher level languages available, or trying to roll your own cryptography algorithms when virtually uncrackable and objectively safer encryption standards exist.

besides, your adverseness is probably a bit misplaced, no? After all we code things for the sake of automating tasks for speed and without human error (assuming we prove within reason that are algorithms are correct)

How often do you query search engines or programming resources? Would you agree with the statement that ~80% of programming is knowing how to Google? At minimum, asking ChatGPT or an equivalent LLM is like using StackExchange on steroids...and surely you've found StackExchange useful in your technological pursuits.

Ultimately, technology is about being efficient (lazy) and reuse (cheating) so we can move on to accomplishing bigger things

I'm doing fine so far. so there is no problem to be solved by a LLM. AI won't get me a girlfriend. AI won't solve my waking up issues. AI won't improve my diet. AI won't make me less sick. AI won't remind the local cafeteria employees they forgot my order. people of the past, in the past few millennia, ever since the dawn of humanity, accomplished a lot without computers.

I'll try my luck. and I'm very lucky.

I don't have this problem. most IDEs have Vim mode, so I'm happy.

I take existing technology for granted and I don't question them, but if I'm allowed to invent an assembly-like DSL to generate code rather than depend on features from higher-level language, I don't hesitate.

that's indeed stupid, I agree.

yeah

all the time, unless there is good documentation

yeah

sure it is useful, sometimes

true. but this doesn't entail LLMs exclusively. other technology fulfills the same purpose. unit testing, fuzzers, end-to-end testing, IntelliSense, code cleanup, refactoring tools, machine-readable formal specifications, ever smarter compilers, sanitizers, static analysis, code-smell analysis, better languages, better libraries, better frameworks, better build tools, syntax highlight, code search engines, formatted comment documentation etc.

all tools that employ thousands of rules handcrafted by human experts over several years, 100% correct and deterministic, no conversational AIs involved, perhaps only classic AI algorithms at best.

I'm still waiting for the day semantic AI stops hallucinating syntax and actually understands context instead of just predicting it with a poker face. That said, using it to translate raw math notation into Mathematica syntax? That’s next-level — borderline wizardry for anyone who's ever had to fight with Wolfram’s parser at 3 AM.

Claude can do it from image recognition of--for example--a blackboard. It can also convert to LaTEX. Seriously, the error rate is better than when I was a freshman and my university still budgeted for secretaries. The lack of tits is definitely a downgrade (Miss Watson, we all miss you), but still a net win.