Yes, it’s extremely powerful. For example I recently had an idea for a script to merge cbz files into a single document so I wouldn’t have to clutter my ereader with many individual chapters. The LLM had almost no issue writing the whole thing with only one revision, spent less time and thought on that than I did googling around to see if there were existing solutions. It’s really nice being able to create programs like that just on the high level concept and without mentally getting into the weeds of implementation details. If I wrote it myself I would have had to refresh my memory on stuff like regex and sorting syntax and it would have been way more time and effort. It basically lets me write custom scripts for any trivial problem where they could be useful where otherwise it might be too much trouble.
Yes and no. I compare it to a graphing calculator: I know how to graph a parabola by hand already, but I don’t want to have to do it over and over already. That’s just busy work for me.
LLMs are similar that way. There’s often a lot of boilerplate to get out of the way that’s just busy work to write over and over again. LLMs are great at generating some of that scaffolding.
LLMs have also become a lot more helpful as Google search has gotten worse over time.
It’s sometimes helpful when working with libraries that are not well documented. Or to write some very barebones and not super useful tests if I’m that lazy. But I’m not going to let it code for me. The results suck and I don’t want to become a „prompt engineer“.
Not really. Writing code is the easy part. It’s not the rate limiting step. The hard part is getting requirements out of customers, who rarely know what they want. I don’t need to push out more code and features faster, that would make things into unmaintainable spaghetti.
I might send it a feature list and ask it “what features did they forget?” or “Can you suggest more features?”, or even better – “which features are the least important for X and can be eliminated?”. In other words, let it do the job of middle-management and I’ll just do the coding myself.
Anyway, ChatGPT blocks my country (I’ve confirmed it’s on their end).
Mostly as a search engine. I have it set up to only respond with answers it has web sources for. Code completion like Copilot can be useful, however 90% of the completions aren’t really saving me any time, the other 10% are awesome though.
So I could easily drop copilot but ChatGPT or HuggingChat used like search engines are awesome.
Depends on if you want to work with existing code. LLMs tend to be good at generating small code snippets but not good at understanding / finding errors in existing code
I’m a new DM (and new to TTRPGs in general). I’m using bard and chatgpt to keep track of homebrew stuff.
I’m running an almost completely custom system, adapted to ASOIAF. Races (renamed to origins), classes, backgrounds, feats, etc. extra mechanics like duelling systems and large battle simulations, and faction interaction systems. It’s a lot, and I find it easier for me to have the bot spray solutions to whatever issue I run into, then grab the one that might work, and refine it until it might sound fun. I need to get a system in order to keep track of my campaign, though. Tried WorldAnvil and honestly, I don’t need that many tools. Might go back to Notion and keep track of all the factions and characters that way. Gonna be a lot of work though.
Tried WorldAnvil and honestly, I don’t need that many tools. Might go back to Notion and keep track of all the factions and characters that way. Gonna be a lot of work though.
Obsidian has been great for me to keep track of all my worldbuilding notes for Pathfinder 2e
Yes, by far the most useful thing is stuff like API and keyword documentation for poorly documented code. Its literally the promise of self generating docs for tedious shit.
But just consider how long it would have taken me to answer each of those questions just by googling and reading old forums and stack overflow posts.
Much like sitting next to someone with experience, a question that could take me hours to answer on the internet took me only seconds to answer by asking directly. GPT’s responses are still long, so it’s not pure conversational style, but the longer responses aren’t wasted fluff. It’s all relevant to what I asked.
Natural language as a way to query a knowledge base is enormously useful. Especially for something that requires update of existing knowledge as often as tech work.
Natural language as a way to query a knowledge base is enormously useful.
Great post. I want to highlight your sentence above as a key point, for folks trying to come to grips with where and how to use the current generation of AI.
Hmm well for research it can give me good pointers, when I am going into a new field.
For actual coding it’s mostly useless for the moment. It’s not trained to be productive, so it doesn’t know what to focus on and tends to be overly verbose. Its internal model of what’s going on is also quite shaky.
It feels like working with clay, I have to somehow get the code the llm generates into the shape I need. But it’s like looking at a movie at super slow mo, and the clay is too wet and keeps falling apart.
Furthermore, it cannot handle anything more than relatively low complexity code. Sure it can give you a function for drawing a circle. But architecture and code smell are things it doesn’t understand.
So after using it for a year I must say that I don’t use it for actual coding. I use it mostly to get an overview of fields I’m not that much into. For example lately I’ve looked into quantum field theory again, and Rust for the first time. I know it spouts a lot of nonsense but I can still get the gist of it.
I mostly use shell-gpt and ask it trivial questions. Saves me the time for switching to a browser. I have it always running in a tmux pane. As for code, I found it helpful for getting started when writing a functionality, but the actual engineering part should be done manually imo. As for spending money on it, depends on how you benifit from it. I spend about 50c on my openai API key, but I know a friend who used ollama (I think with some mistral derivative) locally on a gaming laptop with decent enough results.
I’ve had the most luck with using ChatGPT for troubleshooting my existing code. I typically tend to lean more towards creative coding, and can provide it with my source code and a casual explanation of the issue and it can often explain how to manipulate things in a way I want.
I’ve relied on it a lot less for code generation and found it to be much more useful as a tutor for concepts that I can rework myself. I haven’t spent much time with Copilot since most of my projects are aiming for an uncommon goal.
Where I’ve found it to be less than useful in code generation is I’ll get caught in a loop where it’s trying an approach I’m not familiar with, so I feed it back the errors I’m getting and hoping it can solve it on its own, but it rarely is able to.
I don’t code professionally, but I’d probably hesitate to use it for anything used in production just based on what I’ve experienced.
I’ve used chagpt and google’s new one, whose name eludes me.
In cases where I absolutely have to write in a language or structure I hate, I prototype in an A-I to speed up the experience so I can stop sooner. It saves me so much time doing something I hate.
Add comment