3 ms·

Honest question: What does GPT-3 add to my tookbox as a developer?

powerhour 4y ago
I wonder if anyone has successfully used it to create library documentation. Obviously you'd have to tweak whatever output you get but can GPT-3 provide a substantial starting point?
- guipsp 4y ago
  One thing to note (without commenting if this applies in this case or not) is that sometimes a bad starting point is worse than no starting point
- ianbutler 4y ago
  I've used it in a previous tool to generate documentation from code snippets, it works out pretty well.
- moffkalast 4y ago
  Yes, there are several VScode plugins for it. Not sure if they use GPT-3 specifically or one of the slightly smaller versions though.
simonw 4y ago
We're still figuring that out. It's a really fascinating piece of tech, and has all kinds of non-obvious applications. I wrote about using it to explain code (and mathematical formulas and suchlike) a few weeks ago: https://simonwillison.net/2022/Jul/9/gpt-3-explain-code/ https://simonwillison.net/2022/Jul/9/gpt-3-explain-code/ I've been experimenting with using it to build a human-language-to-SQL tool, so that people can ask questions of their data "what country had the highest GDP in 2019" and it will turn them into the correct SQL query, given a table schema. I'm still iterating on this but it's shown some very promising initial results. I use it a lot when I need to get something small working in a language that I don't have day-to-day familiarity with. "Write a bash script that loops through every MOV file in this folder and extracts the audio as MP3" is a good example of that kind of prompt. Riley Goodside on Twitter posts weird and interesting new things you can do with GPT-3 on a daily basis: https://twitter.com/goodside/ https://twitter.com/goodside/ - his instructional template trick was on HN the other day, it's really clever: https://news.ycombinator.com/item?id=32532875 https://news.ycombinator.com/item?id=32532875
- intrasight 4y ago
  That is the type of application that I am also interested in. But how does one "train" GPT-3 on your business schema? How does one train it on any custom domain?
ivalm 4y ago
Maybe not for a developed, but for an AI based startup: 1. Generate synthetic data that is well aligned to your needs. With careful prompting + ensembling + after-fact human filtering you can generate a lot of very particular human-like data that you can then used to train/etc your product. 2. Generate labels. gpt-3 can give pretty good NLU results through appropriate prompting. You can do multiple prompts + ensembling to get very good labels on free text (sentiment, entity linking, intent, etc). In both above use cases you can actually avoid deploying gpt-3 as part of client facing product, but instead leverage gpt-3 to train smaller "on-rails" models/rules/etc.
thorum 4y ago
GitHub Copilot is built on GPT3.

I wonder if anyone has successfully used it to create library documentation. Obviously you'd have to tweak whatever output you get but can GPT-3 provide a substantial starting point?

guipsp 4y ago
One thing to note (without commenting if this applies in this case or not) is that sometimes a bad starting point is worse than no starting point
ianbutler 4y ago
I've used it in a previous tool to generate documentation from code snippets, it works out pretty well.
moffkalast 4y ago
Yes, there are several VScode plugins for it. Not sure if they use GPT-3 specifically or one of the slightly smaller versions though.

simonw 4y ago

We're still figuring that out. It's a really fascinating piece of tech, and has all kinds of non-obvious applications. I wrote about using it to explain code (and mathematical formulas and suchlike) a few weeks ago: https://simonwillison.net/2022/Jul/9/gpt-3-explain-code/ https://simonwillison.net/2022/Jul/9/gpt-3-explain-code/ I've been experimenting with using it to build a human-language-to-SQL tool, so that people can ask questions of their data "what country had the highest GDP in 2019" and it will turn them into the correct SQL query, given a table schema. I'm still iterating on this but it's shown some very promising initial results. I use it a lot when I need to get something small working in a language that I don't have day-to-day familiarity with. "Write a bash script that loops through every MOV file in this folder and extracts the audio as MP3" is a good example of that kind of prompt. Riley Goodside on Twitter posts weird and interesting new things you can do with GPT-3 on a daily basis: https://twitter.com/goodside/ https://twitter.com/goodside/ - his instructional template trick was on HN the other day, it's really clever: https://news.ycombinator.com/item?id=32532875 https://news.ycombinator.com/item?id=32532875

intrasight 4y ago
That is the type of application that I am also interested in. But how does one "train" GPT-3 on your business schema? How does one train it on any custom domain?

ivalm 4y ago

Maybe not for a developed, but for an AI based startup: 1. Generate synthetic data that is well aligned to your needs. With careful prompting + ensembling + after-fact human filtering you can generate a lot of very particular human-like data that you can then used to train/etc your product. 2. Generate labels. gpt-3 can give pretty good NLU results through appropriate prompting. You can do multiple prompts + ensembling to get very good labels on free text (sentiment, entity linking, intent, etc). In both above use cases you can actually avoid deploying gpt-3 as part of client facing product, but instead leverage gpt-3 to train smaller "on-rails" models/rules/etc.

thorum 4y ago

GitHub Copilot is built on GPT3.