Any advice or experience installing a copilot extension for Visual Studio Code?
Has anyone tried either Github Copilot or any other copilots in the available extensions. Has Particle any plans to list a Device OS trained Copilot? There is a lot of best practice and tutorials as well that could be a lot more consumable if available as an 'AI pair programmer'.
@armor ,
I have been using it. It is fairly helpful as it gets all the tricky syntax bits right. It can also be a bit scary when it correctly guesses what I am trying to do with a function. I believe I am more productive when using it.
I would suggest you give it a try. It is only $10/month and you can cancel if it does not earn its keep. One thing to keep in mind, these things are getting smarter at a breath taking rate.
One issue that I am sensitive to as my "day job" at Red Hat has a generative AI tool for coding automation that is more professionally oriented - attribution. If you are creating commercial code, you should be concerned about where the code that co-pilot suggests has come from. This is important as Co-pilot crawls GitHub and the interwebs gathering the code it uses for its large language model. Co-pilot does not provide attribution for the code it is recommending. Not sure this is an issue for you but, I think it is something we will all hear more about.
Thanks,
Chip
Thanks Chip,
I am aware of the attribution issue for AI generated code. I thought the issue with GitHub copilot was that the training was on public repo code up until mid 2022.
How up to date is it on the Device OS API?
I was hoping that someone from Particle might answer that they have trained a bot on the reference documents and tutorials?
I’ve been using git hub copilot extension for VS Code as well. It’s shocking how accurate the code snippets it generates. I understand that it’s also taking into account all open editors/tabs within your VS code extension which makes it even more impressive what it’s ghost writing for you.
I also appreciate the chat bot right within VS Code. It’s a better end user experience than copy/pasting between chat GPT and VS Code. Most times I can just highlight a code snippet and ask it something.
In my opinion… it’s so well worth the $10 a month. I do a decent amount of SQL work as well… I’m waiting for the day where it can take in the entire SQL relational database schema in generating SQL queries, etc. many times it’ll suggest fields or tables that don’t exist in the DB. I’m sure all in time and as chip said, might even be tomorrow or next week how crazy fast this technology is advancing.
Jeff, thanks for your experience. I am going to try it out. Went to Microsoft Build AI Day event in London last week and was blown away. There is a new thing called AutoGen coming which is a group of AI bots that work together to break down a problem into chunks that specialist bots solve and they iterate until there is an acceptable answer to return!
Curious that no one from Particle has answered? @zach
Just back on this thread having spent some time using Github Copilot on the Particle Workbench.
I was doing 2 things; extending an HTML/Javascript web page using the BLE API to enable setup of P2/Photon2 WiFi. I have not spent any time writing HTML and Javascript so the Copilot suggestions and explanations of code were fantastic. However, it did suggest some code with imaginary vars at one point. Writing a README.md was fantastic - 90% done.
With the Particle sketches I have tried - the suggestions and knowledge is limited to the codebase it can see. It would be really great if the Device OS API documentation was available in a form it could access (vector database?).
Any thoughts from the community about this?