I saw an Information Graphic today from Microsoft that amazed me. I wasn’t sure exactly what it meant, but couldn’t imagine any version of what it might mean that would not be revolutionary or more. It’s even more amazing when we notice that the statistic is more than two years old!
The citation is here:
https://github.blog/2022-06-21-github-copilot-is-generally-available-to-all-developers/
Fortunately, #PerplexityAI was able to give me perspective on this:
It’s important to note that this statistic is specific to Copilot users and not representative of all code on GitHub.
https://www.perplexity.ai/search/what-fraction-of-new-code-is-w-GUdeOhqKQgiT.BCG5pfoeA
Which makes a lot more sense to me.
I still wonder exactly what it means. I am a habitual Microsoft Copilot user for #python, #typescript, and #javascript.
I recently looked at my recent javascript code and tried to do some counting. I figure that I have written less than 25% of the characters. On almost every there was some level of completion (I don’t know which was copilot and which was something else). There were a few complete functions written, lots of near repeats of previous blocks, many wrong guesses, and several things where I would have had to consult the documentation, but almost nothing I would count as having required thought. Less than 3% of what it wrote might count as requiring thought or creativity.
When I first showed off some of my Dall-E creations, some said they didn’t want AI to generate art and poetry for them so they would have more time to do the dishes and the laundry, but rather AI to do the dishes and the laundry so that they could spend more time painting art and writing poetry.
#Copilot does a lot of the dishes and laundry for me (even if more than occasionally takes it more than one try.)
I am realizing that this is how it should be, and the more I think about it, the more grateful I am for the hundreds and hundreds of little things it did in my last thousand lines of code.
