LLMs... Do you use them?

[devils]

AI is the hot topic of the decade, or at least it seems to be so far. Regardless of whether we like it or not, it's here to stay.

Back when I was working in a tech office, I had to use ChatGPT to generate SEO garbage that, while not incorrect (I took the time to proofread and fact-check each text, something my boss wasn't too big of a fan of because I was spending too much time doing it), was basically the same recycled text that appears in every other corporate blog. Not my proudest moment, to be honest. According to my former boss, I'm «really good at using ChatGPT» (whatever that means).

All that to say, I've used LLMs before. In my free time, I've tried to get ChatGPT to hallucinate and I've used DeepSeek to help me find articles about specific topics now that most search engines are being overflown with SEO trash (that I helped create... sorry!). I don't use them frequently at all, but I've found that they can come in handy when I just can't seem to find something very specific.

I won't deny that there are many issues with modern AI usage, but I've been fascinated by the topic for a while now. So, have you used LLMs? If so, which one(s) and why?

[GideonWilhelm]

I have attempted to use it as a shortcut for learning programming in the past, but discovered quickly how easy it is to fall down a rabbit hole and lose your ability to think about the problems you face on the subject.  I ultimately didn't learn anything at all.

Since then, I have occasionally used it as a sounding board for general ideas, particularly because it comes back with questions I hadn't thought of.  I do know it's better to share ideas with friends and family, and I do, but it's always helpful to have someone or something poking at the ideas to see what falls through the sieve.

As to which ones and why, pretty much just ChatGPT free tier.  I tried setting up a local model once and really screwed up my linux install ^^;

[KrazyKoen]

I use Deepseek very occasionally when programming, but only when I'm genuinely stuck and I always make sure I actually understand the code before trying it. I don't want to overly rely on an inherently unreliable technology. I see way too many people trying to use AI like a search engine and it genuinely kind of scares me.

[crazyroostereye]

I usually use them when I have to deal with Boilerplait I can't be bothered of writing at the moment, or when I am really stuck and traditional researching has brought me to a Dead End. Tho in those cases the AI responses are very hit or miss.

But I am using it less as it isn't giving me many Benefits, and honestly I have more Issues at this Point when coding. My biggest use case is when I need a snippet of code I could check quickly and can't be bothered writing myself.

[AY38910]

NEVER. Unless I need a text so ridiculously absurd no human could write it, AI is usually very good at that.

[candycanearter07]

I'm with AY, the whole thing feels wrong to me, even the thought of using it for "inspiration" icks me out because it feels like a massive slippery slope to not making stuff again, and every single big tech problem seems to tie back to it nowadays.

[splashy]

I use ChatGPT every so often when I'm not having any luck with search engines, or just want a summarized overview of a topic. I'll also occasionally use it to slop code when I just want something to work without going through the process of programming.

[Eunice]

I use Bing and Nightcafe to make pretty pictures. My husband has developed an odd relationship with an AI program! I think it's ChatGPT but he also uses a few others. They've helped him to knock his CV into shape. One of them is currently having nice chats with him about Space 1999!

[dim]

i used to use it to help me with javascript, but i realized that, as others have pointed out, it was teaching me nothing and served only as a shortcut - it started feeling like cheating. in the past, i've also used it sometimes as a sounding board about various topics in my life.

nowadays, i've stopped using it completely and have been actively discouraging everyone in my life from doing so, because i've become educated on the horrible practices and principles of the industry running it. i would like to stay as far as possible from supporting the AI industry as it is now, and find myself wishing there were protests against it that i could join.

[Corrupted Unicorn]

Been forced to try ChatGPT due to circumstances outside my control (aka the courses I was taking included them in their assignements) and damn, now I know first-person why it's "the new revolutionary technology" 

I was even more floored when I asked it to do a little creative thing (nothing mine! I don't trust it with my personal ideas  nothing from anybody else either. just a dumb thing). It could string ideas so much faster, and apparently so much better than me.

I admit, I am afraid to write about this because of very extremist sentiments in favour and against LLMs. But at the same time, I was tired of this being a black and white issue (how ironic of me), a situation with only two sides: you either side with the artist-hating AI troll-bros or you side with those who believe that even just opening ChatGPT permanently corrupts your artistic integrity. I wanted to make my own opinions, and to do so, I had to experiment on my own.

So yeah. I got trapped for a while. But I learned a lot of things about LLMs first-hand:

   Its "magic" relies on being able to bring "tailored answers". It'll adapt to anything you give it. This is also why people like to use it instead of search engines because they've been so shit lately, the illusion of personalized answers is way more alluring.

   Its got a "dark magic" aspect to it too. It'll never get tired, or bored, and unless you ask it to, it'll never refuse your ideas or tell you they're stupid  It turns into an instant best friend that's known you and your ideas for long by default. Of course, it's always much more desirable to bounce ideas off with actual people, but if you're a lonely person, I dare you PM the closest Melonland forum member available and tell them "hey want to hear about this cool story I'm thinking with my OCs?". And see how that goes. And if you think that's too easy, now try it in real life    I've seen stories of people getting withdrawn from real life and real people in favour of LLMs to the point of some wanting to marry them  Not to mention, it's intrinsically addicting, not just because it can adapt itself to your input, but because it keeps offering suggestions to keep the conversation going.

   It tends to answer the total "average" when it comes to creativity. If you ask it for a story, it'll give you an average story structure. Sometimes you need the average as a starting point, but relying on it too much doesn't seem like a good idea 

   I even learned to set up a LLM that runs from my own computer. (Owned and tweaked by me! No privacy concerns!). It's nowhere near as good as ChatGPT, but I guess that's because I'm not good at this whole parameters thing AND because ChatGPT has a bigger database and more servers to back it up. But I'm surprised the technology to have "ChatGPT at home" is closer than I thought.

   Eventually, I realized that using a LLM too much can devolve your creative work into a slot machine. Just inputting over and over for a satisfying result or a gacha game. I hate it. I love creative work because it gives me control over something: I get to choose what to draw, and how to do it. This takes it away from me.

   It also made me appreciate my own way to think about things. How I approach my own creative work. How I think of things nobody else has thought of. The LLM does things I've never thought of, because my ideas are too hare-brained to consider the "average" in the first place; but at the same time, the LLM would never come up with my ideas on its own.

   My conclusion? This thing is like jumper cables. If you're stuck with a massive creative block, you can use it to blast it away. But if you try to run your creative engine by jumper cables alone, you'll end frying it up 

[BlazingCobaltX]

No. I understand they can be really useful for certain situations but the environmental cost is so grand that I can't bring myself to rely on them. I'd be more likely to use such models if 1) they were trained using ethically-sourced data (either people consenting to their data being used or someone training a model only with their own data), and 2) they were much more energy efficient.

There was a time I did translations for this company that built their own customer service bot, so I've seen that these things are perfectly possible. The way so many people have jumped onto the Chat-GPT bandwagon with little regard for its ethics is so concerning to me. The fact that none of these LLM developers have to deal with the consequences of their unethical business is equally concerning.

EDIT: When GPT was first released and my friend had it on his phone, I tried two prompts. One was to explain Caspi and Moffitt's theory on antisocial behaviour. It spat out something that was correct, but not necessarily anything more insightful than a second-year student would produce. For full-time research I didn't see it doing more than I could do on my own.

The other prompt was to write a romantic story between Caspi and Moffitt. It was admittedly amusing, but even then nothing special.

[MediocreMetastasis]

I use for job interviews that use AI to review them e.g. sapia.ai. If they're not gonna put in the effort, I won't either. And I always progress to the next interview stage.

I also use it to review and sometimes edit what I write as I am, with a medical reason, abysmal at writing.

I also use it as a debugger as, when it works, it's pretty good... If you know what to ask it. Instead of saying "fix it" say what part of the code you assume is broken.

I mainly only use AI as a tool and I never really use it when I want to look something up (tho sometimes I just use google AI answer and face the consequences of it being misinformation 50% of the time).

[grubbyfox]

I've used it a couple of times for stuff that isn't really something I can "google" nor ask my friends, lmao. That sounds crazy, but it's often about either grammar/a word I cant remember. And asking my friends is very 50/50. Half the time they dont understand what I mean, and I end up trying to feverishly explain on numerous chats what word I am trying to remember/see if exists/get a synonym for without remembering another word for it. A lot of the time it's not an english word either so I cant go on reddit or whatever and ask either (and i dont want to wait 5-7 business days for someone to just auto-delete or downvote my post on reddit as per usual).'

with chatgpt (whatever version really, i dont care, I just chose one that is no-login-required.) - it's really helpful for those things. I can describe what word I am looking for, in what setting it would be used, what vibe it gives/what it sounds similar too, anything really, and it ALWAYS give me what I am looking for, no nonsense. It was also very helpful when I needed some tips for a few acronyms and I've never been good with those. It would have honestly taken me months to figure out some acronyms for some stupid hobby I was trying to enjoy, meanwhile I could ask the chatgpt, based on some requirements, and it came up with some really fun and helpful ones.

I also dont like asking friends for help with words because sometimes I will be using the word in private fiction or something and I dont need them to be like oooh whats this for blah blah.

I've also used chatgpt to figure out some text that was in a broken file. I had some files salvaged off an old harddrive, and some files seemed corrupted, but i could make out a few logical words. I asked chatgpt if it could analyze what had happened with the file (text) and maybe put together anything that looked like readable text/a convo. It managed to tell me what the file most likely had been/what it seemed to hold of info, and it was indeed and old chatlog, which the LLM also neatly put together for me in a readable chat-exchange format with person-A and person-B.

So yeah. I try to not use chatgpt or any AI stuff and I can easily go months without using it. I dont like the idea of it using so much power/eletricity on my stupid whims, so I try to keep it to a minimum. I always look online/try to figure it out myself/ask friends as the first go-to. If that doesnt work or I dont want to/cant do that for various reasons, I consider using AI to help me out.

As for AI images and stuff, I tried it out back in the will smith-spaghetti days to put it that way. I used the one called something like crai-on or something? I dont remember. Very basic and gave mostly just garbled stuff. I needed a bunch of avatars for a fake forums ("art project" thing) and I did not want to use trademarked images and stolen stuff, so I had some random like 50x50px images conjured up, of like, "a car"or "a drink at the beach" or "a cup of cocoa" and stuff like that. i didnt want to incorporate ai stuff, but I decided it was the better option than to trawl real forums and steal people's avatars. Maybe it was a bad decision, but idk. It is what it is, haha.

[devils]

OP here! I've read all your wonderful answers and I'm really pleased with the responses y'all wrote I noticed a lot of you prefer to use AI in a completely different way compared to me. I mostly use it when I need to research really niche topics that don't necessarily show up in most search engines, while a lot of you have used it to get ideas or answers to certain coding questions (which is actually something I prefer doing on my own).

I'd like to add a couple of points to my original post...

• DeepSeek is not available in every language and as such researching for country-specific data can sometimes be a challenge with it. In those cases I use ChatGPT, which is a lot more versatile in that aspect.
• I do agree that there are myriads of problems involving the consent of creators to generate anything including their work... I always ask for sources when I research using it, and I suggest you do the same.

Moving on to a related topic, AI image generation. Eunice mentioned using it, and I'd be lying if I said I never used it myself. However, I am deeply disappointed with its current state... I used to love it when it was janky and you could tell it was clearly made by a machine, like in this image:



Nowadays it seems AI images are all the same cookie-cutter BS that a human could have made better. It's actually a bit upsetting for me to to think about!

Anyways, Corrupted Unicorn's reply was especially fascinating to me and I would like to reply to a very specific point...

  It tends to answer the total "average" when it comes to creativity. If you ask it for a story, it'll give you an average story structure. Sometimes you need the average as a starting point, but relying on it too much doesn't seem like a good idea 

It seems «average» because it is average. This is due to the nature of AI in itself- computers can't think (not yet anyways!), they can only predict what comes after each iteration (in this case, each word). The more data it gathers that has the same structure, the more likely it is to replicate it. It's actually really interesting technology and I hope I'm explaining it well enough, ahahah.

And yes, it's not good to rely on it too much. It's not the worst thing for your brain that can be found online (I'd argue short-form videos are currently the biggest threat to our brains), but it does get most people to be lazier and seek immediate answers to everything. It's important to fact-check each answer not only because the LLM can be wrong, but because it keeps you curious- curiosity is actually very important, especially nowadays that so much misinformation is spread!

Also...

I use for job interviews that use AI to review them e.g. sapia.ai. If they're not gonna put in the effort, I won't either. And I always progress to the next interview stage.

This one is just hilarious. If you can't beat 'em, join 'em I guess...

[st3phvee]

I don't use LLMs at all. I'm soooooo tired of seeing ChatGPT-generated writing and AI-generated images all over the internet. I have misophonia (meaning I get irrationally upset when I hear certain sounds, like people talking with their mouths open, and have to leave the room or drown them out) and I think I might have a LLM version of that particular affliction as well. AI-generated stuff is very obvious to me, and it drives me nuts when I see other people congratulating the "author" or "artist" on their "excellent writing" or their "artistic skills." I don't care if the people who use them are upfront about it, but it rubs me the wrong way when I see them taking credit for something they obviously didn't create.

I also have major issues with the companies behind these products. The way they stole basically everything on the internet, the way they exploited low-wage workers in the Global South to "safeguard" the tech for widespread consumption... If anyone's on the fence about it, you should read Karen Hao's excellent book, Empire of AI.