Ask Lemmy

I am a student in CS, and today my uni has completely changed its policy on AI. They used to completely ban it. Now, professors use LLMs to make their course materials, generate illustrations etc. They encourage using LLMs to understand courses, to work on our projects… I have one practical session where they straight up give us the prompts to use if stuck.

Now, I might get hate and I understand it, but they are truly useful. A course that would’ve taken a whole weekend to understand is all wrapped up in a few hours. You can ask very specific and "niche" questions it’ll understand and explain in the words that are most adapted to you.

Yes, I use LLMs. I do not condone generating "art" from it like images, videos or music.
But it’s such a great tool and a teaching mentor. If you don’t use it to do your homework for you but to have a better understanding of topics you’re not familiar with it, I find it valid.

never, almost everyone who uses it become kinda lazy themselves, and they always keep referring to "chatgpt as an answer to your question"

I have used AI to make a few games for my kids, and a couple of apps that I wanted for my own wants/needs. In both cases it was very frustrating and I cant beleave people that say the current state is 'great'. It was barely fictional and needed constant over-site - if I didn't know at least basic C++/PHP/javascript and html/css I don't know if it would be that usefull.

It helped me to rapidly prototype, but needed lots of work to keep it on track. I can see how agents go rogue and delete whole dir trees etc.

As for the environmental costs, even with out LLMs I think we are fucked with what we have done up to now and what the US seems hell bent on bringing on us (the rest or the world)

I read that they're not terrible when used to power NPC's in games.

https://www.pcgamer.com/software/ai/company-that-makes-generative-ai-powered-npcs-reports-that-95-percent-of-players-enjoy-their-generative-ai-powered-npcs/

Not my personal take, mind you, but thought it relevant.

I used Copilot to build me a performance review based on actual data (which I reviewed and edited) and my boss said it was the best one he received from 30 people on the team.

Ever? Sure.

I think it's useful as a starting point for a lot of things. I can ask AI a question about a topic I know nothing about, which will typically give me some context on the topic and the terminology to do further research.

At work I do not think their use is ever justifiable because the rapidly increase the amount of satisficing behavior from my colleagues. I have had many experiences where obviously bad work was submitted that was clearly llm generated and it was clear that the person submitting it just generated the output and handed it over. People turn their brains off.

The other thing I have noticed about their use is, once you start caring about the quality of your work, their value plummets. If I were to use one for my work I would need to check its output by experimenting with code, doing research, thoroughly considering both sides of an argument, etc. But if I were not going to use one I would do my own work by experimenting with code, doing research, thoroughly considering both sides of an argument, etc. So what is the advantage to using one? Either way I am still putting forth the effort to ensure my work-product is high quality. Going clackity-clack on the keyboard is not the hard part of my job, all the other stuff is.

Probably programming, i used Claude Code recently and it's wonderful because i can use it for debugging stuff that i can't undersfand or i can make him do boring ass stuff that i could do myself but just eat my time uselessly (So i can go make a mug of tea or a cup of coffe! Not work XD)

But i want to underline that it will NEVER replace a programmer, to stupid to do it rn, but it can really help our work and that we need to push for self-hostable AIs, right now there are many models but they either require too much resources or are stupid...i hope that we improve their efficency more...

I agree that they are unethical rn because it's all stolen stuff, i hope to be able to make my own AI in future that is trained in a more ethical way(just as hobby open source project tho)

I think its great for inspiration but your final product should never be raw AI/LLM output

No one wants this beside for people who lack creativity.

What AI should be doing is learning how to take out my garbage, cut my grass, and do the dishes for me. Not whatever this dystopian bullshit is.

I feel like self hosting LLMs and GenAi is slightly better for the environment, definitely less environmental impact than gaming.

It just these massive datacenters and models. If people can just be a little more patient and specialized with their AI usage it saves so much electricity

If you made all the training data yourself, or ethically acquired the training data, then go nuts do whatever you want with it. See Corridor Digital's ai chroma key thingy.

If the training data isn't ethically sourced, then it gets iffy.

I use ai primarily for my own entertainment. None of it are things I'd want to share with the world. Is "dicking around" justifiable? Eh. I eat meat and shop at amazon, both of which are things that some people would find "not justifiable", so someone is going to be upset with me no matter what.

In the case of artistically, I don't take offense to ai tools being part of a process, what's important to me is that the ai isn't the entire process. You wouldn't go to a cinema, record the movie with your phone camera, then post it online saying "look at what I made". That's nonsense. But if you took clips of that same movie, rearranged and dubbed over them thus creating a new unique work, you could post that online and say "look what I made". Whatever the ai output, no matter how detailed your prompt, should be treated as being made by someone else. You don't get to say "look what I made" unless you actually do something with it.

Another use case is summarizing conversations and compiling notes. This is another one that I do often. I could go on for hours about a subject (usually while drunk) and at the end I tell it "compile a detailed report on everything discussed, be verbose and leave out no details" or something similar, and that output goes into my notes documents. It's fine to copy pasta that, because it's not going to be anything that anyone ever actually sees.