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

Command line functions around OpenAI

kadekillary.work

Post reply on HN
Print view View original

201–210 of 316 posts

Re: Command line functions around OpenAI

#201

Post by throwthrowuknow »

juujian wrote: post #33

Don't get me wrong, this is neat. But the only reason why this works is because there is one million tutorial out there who use this data for teaching new students. The bigger issue is edge cases, stuff where there is no precedence. This output is a starting point for something , but it does not even get you 3% of the way there unless you work for a teenager who is into basketball anecdotes.

It’s also trained on code and books. The knowledge is in there, it just needs more RLHF to improve its abilities. There’s a good chance it can handle those edge cases by combining the information it has in different ways.

Re: Command line functions around OpenAI

#202

Post by noncoml »

Am I the only one that doesn’t get it?

I tried GPT and it’s unable to write anything but the most simple code from simple requirements.

As a tiny bit of complexity and it’s lost.

Are people at work dealing with tiny code bases with well defined requirements?

Re: Command line functions around OpenAI

#203

Post by Zetice »

throwthrowuknow wrote: post #196
Zetice wrote: post #185

Ok, dumb question, but does the API skip over the whole, "Thanks for asking that question, here's what I found:" stuff I keep seeing in ChatGPT when I ask it to generate non-text responses? Even when I ask it the "generate a 10 row csv containing sample chipotle restaurant order data with headers" prompt, I get "Sure, here's an example CSV file with 10 rows of sample Chipotle restaurant order data, including headers:…

Add: I only want the CSV, don’t describe what you are doing.

Heh nope, I still get the whole "Sure, here's an example of a CSV file with 10 rows of sample Chipotle restaurant order data:"

I wonder if the web UI has some additional "explain everything you do" prompt injection stuff going on.

Re: Command line functions around OpenAI

#204

Post by macromaniac »

Zetice wrote: post #185

Ok, dumb question, but does the API skip over the whole, "Thanks for asking that question, here's what I found:" stuff I keep seeing in ChatGPT when I ask it to generate non-text responses? Even when I ask it the "generate a 10 row csv containing sample chipotle restaurant order data with headers" prompt, I get "Sure, here's an example CSV file with 10 rows of sample Chipotle restaurant order data, including headers:…

Are you using the API endpoint or browser chatgpt? I tried it 5 times via API and it works and then 2 times via the browser and it doesn't work. They are different models.

If you are using the API try lowering the temperature mb?

Re: Command line functions around OpenAI

#205

Post by hyperthesis »

surgical_fire wrote: post #178
angarg12 wrote: post #151

Earlier quoted context omitted.

I discussed my experience using ChatGPT in HN before and I was accused of being skeptical. But it isn't true at all, I think LLMs are a wonderful tool that will change software development. However I'm not worried about losing my job. As you say, I find the act of writing code is hardly the bottleneck to my productivity. One of the best case scenarios, LLMs become the ultimate high level language. In that case, Softw…

> As you say, I find the act of writing code is hardly the bottleneck to my productivity. When people claim ChatGPT can replace software engineers because it can sort of code is always puzzling to me. 20 years into this career, and writing code is only a minor part of my day-to-day activities. A part I do enjoy doing, but still minor part nonetheless. > What if you design an organization so that coding is the bottlen…

> boilerplate

Design Patterns are said to compensate for features missing in a language (e.g. Visitor for multiple dispatch). ChatGPT is sounding similar - solving accidental complexity that "shouldn't" be there in the first place.

Re: Command line functions around OpenAI

#206

Post by 0xdeadbeefbabe »

nkingsy wrote: post #181
roenxi wrote: post #10

Being efficient at writing code can barely make someone a 5x engineer. A team of people who write code effectively is easy to do. Hyperproductivity in software is all about deciding what problems to tackle. Richard Hipp isn't worried about chipotle restaurant orders in golang, he's worried about how to store data reliably. That isn't a coding puzzle that ChatGTP is likely to help with. Either ChatGTP can do the whole…

I briefly worked with a software engineer who was making about a million dollars a year. I mentioned that an approach would be a lot of work. His response: “It’s just typing. Typey typey type (makes fingers on keyboard gesture).” Absolute lightbulb moment for me. To become a senior engineer is to realize that the code writing part is inconsequential as long as there are no technical unknowns in the work. It’s freed m…

Seplly spell spell too.

Re: Command line functions around OpenAI

#207

Post by blagie »

nolawi89 wrote: post #21
blagie wrote: post #12

Earlier quoted context omitted.

I find ChatGPT is pretty helpful for debugging. A big chunk of debugging is missing stupid things, and for debugging, it's like having a second set of eyes. It's almost useless for high-level design and architecture. Those tasks do take more time than writing code, but they're super-charged by rapid prototyping, which ChatGPT excels at. For example, right now, if I'm picking between three libraries, I can have workin…

Not yet; I spend 60% my day reviewing PRs. I don't have to if chat GPT gets integrated as a PR reviewer..

GPT can't do full code reviews. GPT is very good at finding improvements, such as making code more functional, bad variable names, etc. In some sense, it's like ESLint or pycodestyle on steroids.

It won't find subtle race conditions, abstraction violations, or deeper issues.

I find it makes code reviews shorter and more productive, so perhaps you'll spend 10% of your time instead of 60%. You won't spend 0% of your time, though.

Post reply on HN