Notice: Most of the important questions are asked in the outline. Not all of these questions are asked, though, so feel free to contribute here!
This is just a scratchpad which can be used to bring up miscellaneous questions about Github Copilot.
-
How well does Github Copilot autocomplete essays, papers, and similar projects?
-
How does Github Copilot compare to other sources:
- The internet (post a question on Reddit)
- Other students
- Professionals
-
Should students be able to use Github Copilot on assignments? After all, it will eventually could be part of their workflow. After all, it's just a tool. If they can't, then can autocomplete+tab programs be used? What's the difference? Can students use programs to generate getter and setter methods? How much of their programming should be able to be automated? Is snippets cheating?
- Does this make spell checkers and grammar checkers which suggest better ways to write immoral?
- What if you're just writing simple functions you already know how to make? Can it be used then?
-
How will this and other GPT-3/OpenAI models be used in other classes? Will they be used to write papers and essays for students? What if studnts only use it to help put words to their thoughts?
-
If the tool is just a tool like Github claims, can it be used? https://copilot.github.com/'s FAQ
- Like using a calculator to calculate a derivative in a calculus class - it's cheating
-
Could this be used to force styles? If it's used to write papers, will it be used by malicious actors to subtly change the language of a society?
-
How will this be used in the future?
Is using Copilot really cheating? Where do we draw the line for cheating?
- There's also a tool called Tabnine which is similar. It seems, however, that Tabnine is meant for smaller predictions. Tabnine is based on GPT-2 rather than OpenAI. Which performs better?