Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

ChatGPT Changed How I Write Software

readysetcloud.io

Post reply on HN
Print view View original

11–20 of 51 posts

Re: ChatGPT Changed How I Write Software

#11

Post by skilled »

dimgl wrote: post #2

Yeah I'm not convinced. As usual with most pro-ChatGPT posts I see, there is little to no actual useful information on how to use ChatGPT practically. When I read one of these articles or posts, I usually expect to see concrete prompts, their output, and how they can be used to speed up development. Instead all I see is generic vagueisms; "it helps me write unit tests", "it helps me code faster", or in the case of th…

Yeah the author dragged this out for basically no value at all. Anyone who has done programming for the last 5+ years has figured out ChatGPT inside and out.

Re: ChatGPT Changed How I Write Software

#12

Post by jayd16 »

The comment about asking ChatGPT for the content (in this case an alternative to a workout database) is interesting. They supposedly saved "weeks" not building up the domain knowledge necessary to build up proper workout plans. ...but without doing that work, how do they know the AI content is any good?

Re: ChatGPT Changed How I Write Software

#14

Post by infinityio »

> If you routinely pass in a dataset to AI, it can identify anomalies and watch to see if it settles or even reason why the anomaly occurred in the first place. You can provide the AI some guidelines to only alert when something needs attention if it meets a certain set of criteria. This ultimately increases the effectiveness of your on-call crew and lowers your TCO purely based on manpower.

> ...

> So you could use ChatGPT to identify activity trends of an individual to get a statistically high likelihood of reaching them online. Imagine this audit history

> [dataset of login and logout timestamps] `Return a json object with a "sendAtTime" property in ISO-8601 format that has the highest likelihood of reaching the person the next time they are online.`

This feels like it is slightly conflating the strengths of LLMs and AI as a whole.

There are certainly machine learning models that would assist in predicting when a user will be next online, and using an AI assistant to assist in writing such an implementation would likely be helpful.

However, passing JSON into an LLM, asking it to emulate a program to perform this functionality, and then parsing the response to extract other JSON doesn't feel like a stable approach (for example: I was unable to recreate their successful execution under ChatGPT v3), especially given this technique, if widely adopted, may end up including user-controlled fields that would make the system vulnerable to prompt injection.

Re: ChatGPT Changed How I Write Software

#15

Post by jstx1 »

dimgl wrote: post #2

Yeah I'm not convinced. As usual with most pro-ChatGPT posts I see, there is little to no actual useful information on how to use ChatGPT practically. When I read one of these articles or posts, I usually expect to see concrete prompts, their output, and how they can be used to speed up development. Instead all I see is generic vagueisms; "it helps me write unit tests", "it helps me code faster", or in the case of th…

GPT-4 is considerably better than the free version. And I find it hard to believe that whichever model you used has been wrong 100% of the time.

Re: ChatGPT Changed How I Write Software

#16

Post by danjc »

dimgl wrote: post #2

Yeah I'm not convinced. As usual with most pro-ChatGPT posts I see, there is little to no actual useful information on how to use ChatGPT practically. When I read one of these articles or posts, I usually expect to see concrete prompts, their output, and how they can be used to speed up development. Instead all I see is generic vagueisms; "it helps me write unit tests", "it helps me code faster", or in the case of th…

It helps if you don't know what you're doing. If you know what you're doing it just adds friction.

This is actually not intuitive but should not be surprising since the perfect specification for code is well... code

Re: ChatGPT Changed How I Write Software

#17

Post by jchw »

calmoo wrote: post #3
dimgl wrote: post #2

Yeah I'm not convinced. As usual with most pro-ChatGPT posts I see, there is little to no actual useful information on how to use ChatGPT practically. When I read one of these articles or posts, I usually expect to see concrete prompts, their output, and how they can be used to speed up development. Instead all I see is generic vagueisms; "it helps me write unit tests", "it helps me code faster", or in the case of th…

Have you tried GPT-4? Its at least an order of magnitude better than 3.5,and has rarely produced code that doesn't compile for me. The difference is night and day.

My problem with GPT-4 so far is that it's most impressive at doing things that I don't really need any help with, but when I ask it something that I find genuinely challenging, it's error prone and often generates suboptimal code. Of course, it's an extreme far cry from Markov chains and earlier models, but at the end of the day, I think it's skipping too many steps right now.

To be clear, it's still really impressive. Being able to ask an AI assistant to implement e.g. a Fourier transform in Rust and getting an answer that is correct or quite close is pretty damn impressive. That's something that a lot of programmers would struggle with. But, in a lot of other ways, GPT-4 feels very skin deep, and the illusion is pierced when you realize it has just made a bunch of shit up in its impressive-sounding answer.

Re: ChatGPT Changed How I Write Software

#18

Post by unshavedyak »

darthbanane wrote: post #8
dimgl wrote: post #2

Yeah I'm not convinced. As usual with most pro-ChatGPT posts I see, there is little to no actual useful information on how to use ChatGPT practically. When I read one of these articles or posts, I usually expect to see concrete prompts, their output, and how they can be used to speed up development. Instead all I see is generic vagueisms; "it helps me write unit tests", "it helps me code faster", or in the case of th…

I'm in the same boat I have yet to find an actual use case for chatgpt beyond a google/stackoverflow search that uses the same variable names as my code. In several occasions it has produced code that took me longer to fix than just reading the documentation and in almost all of those cases it couldn't help me find the issue and I actually had to read the documentation anyway (it took just seconds to pinpoint the mis…

I use it (GPT4) to explore ideas. Come up with names, explore domains i'm unfamiliar with, etc - all while assuming everything is wrong and i need to validate everything. Which is basically the same as Google/etc. I just use it as a frontend for Search, generally.

I don't use it to write anything concrete, i have not found it to be good at that. I also use FastGPT (Kagi's GPT) as a .. faster search frontend. FastGPT fails to find stuff frequently but it also is so damn fast it feels faster than me searching+clicking_link+skimming_page.

I also like ChatGPT for exploring APIs which i may not know, but can easily validate. It's wrong often, but when i can validate it's not a big deal, and what i get out of it is the ability to Q&A. I find API documentation frequently frustrating because while i prefer documentation, i frequently don't know how to get to my answer. Documentation often feels like i need to read the entirety of everything before i can reliably get any answer, and frankly i'm impatient. GPT gives me a more interactive exploration of content. It either knows the answer, or can often point me in the direction to find it myself.

All this is to say ChatGPT & friends have not revolutionized my work. Hell during my day job i rarely use it because it's a domain i know well. But i do have the belief that it helps me enough to justify the cost, currently.

Re: ChatGPT Changed How I Write Software

#19

Post by Wowfunhappy »

hayksaakian wrote: post #7
dimgl wrote: post #2

Yeah I'm not convinced. As usual with most pro-ChatGPT posts I see, there is little to no actual useful information on how to use ChatGPT practically. When I read one of these articles or posts, I usually expect to see concrete prompts, their output, and how they can be used to speed up development. Instead all I see is generic vagueisms; "it helps me write unit tests", "it helps me code faster", or in the case of th…

to back up the author: I'll use ChatGPT with the GPT4 model and it mostly just works by simply asking it to do what I want, the same way I might ask another person. The main constraints I see right now are around 1) complex projects that don't fit into the token context, and 2) keeping up a "base prompt background info" to describe the general purpose of your program and the technical stack. Issues I've run into: 1)…

For issues 1 and 3, I usually have success asking ChatGPT to correct them. Ie "Can you write this with fewer callbacks using async/await?" or "Can you make this code faster to execute?"

Re: ChatGPT Changed How I Write Software

#20

Post by mjburgess »

jayd16 wrote: post #12

The comment about asking ChatGPT for the content (in this case an alternative to a workout database) is interesting. They supposedly saved "weeks" not building up the domain knowledge necessary to build up proper workout plans. ...but without doing that work, how do they know the AI content is any good?

They don't.

And when you're an expert in an area, you find ChatGPT presents folk opinion rather than expert opinion.

It's very much a journalist of the internet

Post reply on HN