Hello,
This is my first post on Lemmy! Glad to join the Fediverse.
I’d like to ask for advice from people who might have tips on how to do research and learn things without using AI.
I’m starting to get sick of it, it’s very addictive, and I must have spent over €50 developing crap on Cursor.
When I discovered AI in 2023, I was the first person at my high school to discover ChatGPT. In fact, I’d even been playing around with the OpenAI Playground before that.
But over time, I became addicted, it turned into a real drug. I didn’t even use search engines anymore, just ChatGPT, Gemini, Claude, Grok, Perplexity, DeepSeek, and so on…
Since I discovered this in high school, can you imagine what it’s like to discover this while you’re in the middle of a specialized digital studies program? Yeah, you don’t learn anything new, you gradually fall behind. I’ve lost some of my Python skills, like how to create classes and stuff like that, even though back in 2021 I built a client for downloading pirated games that I was pretty proud of. I’m really disappointed in myself about that.
Now, I’d like to relearn computer science, learn how to code, use Linux systems, and so on... but how? Not only have I regressed a bit, but what AI has really brought out in me is a reluctance to do things and a tendency to leave everything up to the AI.
Today I’m in college, and everyone uses it in graded group projects; no one follows the university’s guidelines, which require you to disclose when you’ve used AI. There is a climate of mistrust nowadays.
I’d love to hear your tips, how do you manage without AI, and how can I relearn to live without it? I’ve made up my mind: I don’t want to use it at all anymore, except as a translator (I use DeepL, which I don’t think is the same thing as a generative AI).
You are not the first to ask this here.
Dive into the deep end! Try linux on your computer. Backup your stuff, read up on boot-loaders (and if you want to dual boot Windows at all), then follow the Arch Linux installation guide on their wiki, which is a great "crash course" in linux. CachyOS Linux would be my recommendation for something quicker to install but still hands-on.
As for programming, personal projects will have to stem from something you're interested in. I'd recommend a game mod if you still like gaming, but it just depends on your personal interests.
Also:
This is not your fault. ChatGPT and such are designed to be shamelessly addictive just like Facebook, Twitter and such. Those "AI" apps/services are abominations.
Don't even use the term AI, unless its used disparagingly! IMO its a marketing phrase built to anthropomorphize tech company services, so be more specific. Use "LLM," machine learning, whatever. Basically, use "AI" to mock the tech bro hype.
You might need some more general digital detox. Get addictive crap off your phone. But simply being on Lemmy is a good sign you're on that path!
You need LLM detox. But. In the future, if you are pressured back into usage by a specific class, your career, or whatever, consider the local LLM path, run on your own hardware, hackable, and free of the enshittification of the corporate apps. Consider keeping it offline to force you to fill/use the context yourself. In other words, there is another path besides running back into the drug dealer's arms. But let me emphasize that you should just go cold turkey now, especially at this specific point in your life.
Agreed on all points, but especially the last one. I use local LLMs (bad place to state that maybe?) and find them useful, but just like I wouldn't recommend any level of alcohol to a recovering alcoholic despite it in general being fine in moderate amounts, LLM usage has to be treated the same way.
I'm similarly uncomfortable with the term "AI" by itself. It feels like it's trying to capitalize on decades of fiction and future speculation about AI and take advantage of people who can't know that what we have now is something that falls far short of that. However, I have been using "generative AI" to encompass all text, image, audio, video, etc. generation. I'm curious if you have a preferred term for that without "AI" or if the "generative" is enough of a qualifier to make the term adequately specific?
Yes!
I was afraid to bring it up. I almost didn't. But I just wanted to convey "if your future employer MAKES you smoke, if it absolutely cannot be avoided, be aware DIY nicotine patches exist as an alternative," as it seems like OP hasn't done much with self hosted stuff.
I am just more specific!
For text models, or multi-input models? I use "text generation" or "LLM."
For media generation, I use "diffusion" as basically all those models are diffusion models.
I use "computer vision" for image recognition stuff. "Machine learning" as a more general term.
Basically, I pretend like I live in 2021/2022 before "AI" became a catchphrase, but the field of machine learning was racing away. TBH I hate the term "generative AI" as ALL machine learning is generative in some way, yet old school machine learning isn't normally included in that term.
Thanks for the thoughtful reply. I'll definitely tweak my vocabulary.