I start with "a buggy LLM wrote this piece of code..." then i paste my code for review, so they can shit and bash on it "you're absolutely right: that LLM done a disaster, this is a mess, look how inefficient is this function, here is how it can be improved..."
That's great example of how to get the right answer on any forum. Though you've found away you don't need two accounts. Traditionay you need a archive to give an absolutely complete bat shit answer so a thousand users can tell you why you're wrong.
I usually combine both to unblock myself. Lately, SO, repository issues, or just going straight to the documentation of the package/crate seem to give me faster outcomes.
People have suggested that my prompts might not be optimal for the LLM. One even recommended I take a prompt engineering boot camp. I'm starting to think I’m too dumb to use LLMs to narrow my research sometimes. I’m fine with navigating SO toxicity, though it’s not much different from social media in general. It’s just how people are. You either take the best you can from it or let other people’s bad days affect yours.
If you're planning on using LLMs for coding advice, may I recommend selfhosting a model and adding the documentation and repositories as context?
I use a a 1.5b qwen model (mega dumb) but with no context limit I can attach the documentation for the language I'm using, and attach the files from the repo I'm working in (always a local repo in my case) I can usually explain what I'm doing, what I'm trying to accomplish, and what I've tried to the LLM and it will generate snippets that at the very least point me in the right direction but more often than not solve the problem (after minor tweaks because dumb model not so good at coding)
That's a really cool idea actually. I never considered that you could use such a crazy low quant to, it sounds like, temporarily "train" it for the task at hand instead of having to use up countless watt hours training the model itself!
That's how I use these things, too. Not to "help me code", but as a fancy search engine that can generally nudge me towards a solution I can work out myself.
I do use the 1.5b of whatever latest ollama with open web ui as frontend for my personal use. Although I can upload files and search the web it's too slow on my machine.
If you've got a decent Nvidia GPU and are hoping on linux, look into the Kobold-cpp Vulkan backend, in my experience it works far better than the CUDA backend and is astronomically faster than the CPU-Only backend.
When/If you do, a RTX3070-lhr (about $300 new) is just about the BARE MINIMUM for gpu inferencing. Its what I use, it gets the job done, but I often find context limits too small to be usable with larger models.
If you wanna go team red, Vulkan should still work for inferencing and you have access to options with significantly more VRAM, allowing you to more effectively use larger models. I'm not sure about speed though, I haven't personally used AMDs GPUs since around 2015.
I've been having good luck with Kimi K2 for CSS/bootstrap stuff, and boilerplate API calls (example: update x to y, pulling x and y from this .csv). I appreciate that it cites its sources because then I can go read more and hopefully become more self-reliant when looking up documentation.
Just this morning I asked claude a very basic question, it halucinated the answer 3 times in a row. zero correct solution. first answer It halucinated what a certain cli daemon does, second solution it provided an alternative that it literally halucinated, the thing didn't exist at all, third solution it halucinated how another application works as well as the git repo for said application (A. doesn't even do the the thing Claude describe and B. the repo it provided had NOTHING to do with the application it described) I just gave up and went to my searx and found the answer myself. I shouldn't have been so lazy.
51 Comments
MonkderVierte@lemmy.zip · 260 pts · 320d
WorldsDumbestMan@lemmy.today · 36 pts · 320d
LMAO! Well, at least I can cry to Claude about it...
Aneb@lemmy.world · 11 pts · 319d
This is why I'm on Lemmy. You guys get me
buttnugget@lemmy.world · 5 pts · 319d
This is 100% every normal person’s experience on there at one time or another. Having it be a dupe is perfect.
Petter1@discuss.tchncs.de · 6 pts · 319d
My boiled egg has to be just a little bit liquid on the inside 🌚
MonkderVierte@lemmy.zip · 8 pts · 319d
Petter1@discuss.tchncs.de · 2 pts · 319d
Sticky =/= Liquid
🧐😜💦
MonkderVierte@lemmy.zip · 4 pts · 319d
Corrected.
Petter1@discuss.tchncs.de · 3 pts · 318d
❤️🔥
BossDj@piefed.social · 4 pts · 319d
Downvote harder damnit! I think I broke my screen
magic_lobster_party@fedia.io · 118 pts · 320d
That’s a stupid question it doesn’t deserve an answer. You should be ashamed you even thought about it.
Sonotsugipaa@lemmy.dbzer0.com · 81 pts · 320d
Your description of the problem has words I've heard before, like "a" and "even"; marked as duplicate.
ronigami@lemmy.world · 6 pts · 319d
This has been asked before. The thread was closed with zero comments but suck it.
fibojoly@sh.itjust.works · 4 pts · 319d
Now, that's the real SO experience.
Wispy2891@lemmy.world · 40 pts · 319d
I found a workaround for this:
I start with "a buggy LLM wrote this piece of code..." then i paste my code for review, so they can shit and bash on it "you're absolutely right: that LLM done a disaster, this is a mess, look how inefficient is this function, here is how it can be improved..."
melfie@lemy.lol · 2 pts · 319d
zjti8eit@lemmy.dbzer0.com · 1 pts · 313d
That's great example of how to get the right answer on any forum. Though you've found away you don't need two accounts. Traditionay you need a archive to give an absolutely complete bat shit answer so a thousand users can tell you why you're wrong.
Wynnstan@lemmy.world · 38 pts · 319d
ChatGPT is a narcissists dream app.
lord_ryvan@ttrpg.network · 1 pts · 302d
Yeah, it talks like it was made for HR managers.
pinchy@lemmy.world · 37 pts · 319d
SO: “that’s a stupid question!” GPT: “that’s a great question!”
jumping_redditor@sh.itjust.works · 19 pts · 319d
stack overflow at least is polite enough to call you a moron for asking
Hupf@feddit.org · 2 pts · 319d
https://youtu.be/nhhXBBgdfuc
anton@lemmy.blahaj.zone · 35 pts · 320d
Well, if I asking for help, it's probably because I am wrong about something. So I know who to trust.
finitebanjo@piefed.world · 27 pts · 320d
cupcakezealot@piefed.blahaj.zone · 26 pts · 320d
piefed is working on solution and answer features and i can't wait for stackoverflow like communities without the ai "enhancements"
MonkeMischief@lemmy.today · 2 pts · 319d
This sounds pretty exciting and I keep hearing more and more about piefed lately. I'm kinda excited for this new burgeoning era of the federated 'Net!
goatinspace@feddit.org · 17 pts · 320d
mstrk@lemmy.world · 15 pts · 320d
I usually combine both to unblock myself. Lately, SO, repository issues, or just going straight to the documentation of the package/crate seem to give me faster outcomes.
People have suggested that my prompts might not be optimal for the LLM. One even recommended I take a prompt engineering boot camp. I'm starting to think I’m too dumb to use LLMs to narrow my research sometimes. I’m fine with navigating SO toxicity, though it’s not much different from social media in general. It’s just how people are. You either take the best you can from it or let other people’s bad days affect yours.
Sonotsugipaa@lemmy.dbzer0.com · 35 pts · 320d
Darkcoffee@sh.itjust.works · 18 pts · 320d
They always accuse the user of being the problem when using glorified if-else machines.
xthexder@l.sw0.com · 16 pts · 319d
LLMs have a bit of RNG sprinkled in with the if-else to spice things up.
Darkcoffee@sh.itjust.works · 9 pts · 319d
Have some RNG in your garbage AI, as a treat.
marcos@lemmy.world · 10 pts · 319d
If SO doesn't have the answer to your question, LLMs won't either. You can't improve that by prompting "better".
They are just an easier way to search for it. They don't make answers up (or rather, they do, but when they do that, they are always wrong).
ArsonButCute@lemmy.dbzer0.com · 8 pts · 320d
If you're planning on using LLMs for coding advice, may I recommend selfhosting a model and adding the documentation and repositories as context?
I use a a 1.5b qwen model (mega dumb) but with no context limit I can attach the documentation for the language I'm using, and attach the files from the repo I'm working in (always a local repo in my case) I can usually explain what I'm doing, what I'm trying to accomplish, and what I've tried to the LLM and it will generate snippets that at the very least point me in the right direction but more often than not solve the problem (after minor tweaks because dumb model not so good at coding)
MonkeMischief@lemmy.today · 6 pts · 319d
That's a really cool idea actually. I never considered that you could use such a crazy low quant to, it sounds like, temporarily "train" it for the task at hand instead of having to use up countless watt hours training the model itself!
That's how I use these things, too. Not to "help me code", but as a fancy search engine that can generally nudge me towards a solution I can work out myself.
mstrk@lemmy.world · 2 pts · 319d
I do use the 1.5b of whatever latest ollama with open web ui as frontend for my personal use. Although I can upload files and search the web it's too slow on my machine.
ArsonButCute@lemmy.dbzer0.com · 3 pts · 319d
If you've got a decent Nvidia GPU and are hoping on linux, look into the Kobold-cpp Vulkan backend, in my experience it works far better than the CUDA backend and is astronomically faster than the CPU-Only backend.
mstrk@lemmy.world · 3 pts · 318d
Will look into that when I have some money to invest. Thank you 💪
ArsonButCute@lemmy.dbzer0.com · 3 pts · 318d
When/If you do, a RTX3070-lhr (about $300 new) is just about the BARE MINIMUM for gpu inferencing. Its what I use, it gets the job done, but I often find context limits too small to be usable with larger models.
If you wanna go team red, Vulkan should still work for inferencing and you have access to options with significantly more VRAM, allowing you to more effectively use larger models. I'm not sure about speed though, I haven't personally used AMDs GPUs since around 2015.
smh@slrpnk.net · 2 pts · 319d
I've been having good luck with Kimi K2 for CSS/bootstrap stuff, and boilerplate API calls (example: update x to y, pulling x and y from this .csv). I appreciate that it cites its sources because then I can go read more and hopefully become more self-reliant when looking up documentation.
Kolanaki@pawb.social · 15 pts · 320d
People on Social Media: "You absolutely are an asshole."
cloudy1999@sh.itjust.works · 2 pts · 319d
^^^ Look at this absolute asshole!, ^^^
<--- also this one
dumbass@aussie.zone · 9 pts · 320d
Me: You fucking idiot.
Sonotsugipaa@lemmy.dbzer0.com · 7 pts · 320d
That's just the average stackoverflow comment
copd@lemmy.world · 8 pts · 320d
Answer: Why don't you try searching for the question first?
Me (confused face): How tf do you think I found this page?
Answer: Why are you doing it that way?
Me (sighs): Because I'm an idiot and don't know what I'm doing, is that what you wanted to hear?
Sonotsugipaa@lemmy.dbzer0.com · 16 pts · 320d
tomiant@programming.dev · 8 pts · 319d
User: "Can I ask a q..."
Stackoverflow: "NO!"
b0ber@lemmy.world · 3 pts · 317d
Every second the most ridiculous question gets positive feedback so yeah who knows where we'll end up from here.
stupidcasey@lemmy.world · 3 pts · 320d
I knew you write a function like this f(x) stack overflow has no idea what they are talking about.
rozodru@piefed.social · 3 pts · 320d
it's getting worse too.
Just this morning I asked claude a very basic question, it halucinated the answer 3 times in a row. zero correct solution. first answer It halucinated what a certain cli daemon does, second solution it provided an alternative that it literally halucinated, the thing didn't exist at all, third solution it halucinated how another application works as well as the git repo for said application (A. doesn't even do the the thing Claude describe and B. the repo it provided had NOTHING to do with the application it described) I just gave up and went to my searx and found the answer myself. I shouldn't have been so lazy.
ChatGPT isn't much better anymore.
Itdidnttrickledown@lemmy.world · 2 pts · 318d
And neither can get it right over half the time.
BilSabab@lemmy.world · 2 pts · 318d
Man, I miss those batshit crazy threads about esolangs!
_cryptagion@anarchist.nexus · 1 pts · 320d
sip@programming.dev · 1 pts · 320d
fuck me