Just out of curiosity, is the Vuo Team considering utilizing tools like GPT-4 to develop Vuo, since it’s showing to be such a useful tool for programmers to increase productivity? Maybe there are reservations in using online tools like this for development, what if it were an offline tool?
This is non-programmer me just asking because I’m interested in what the Vuo Team thinks of it. Others are welcome to comment too, just wondering.
While it’s true that some developers have been able to use GPT-4 successfully for some tasks — my coworker mentioned someone using it to generate templates for boilerplate documentation, for example — from what I understand it seems unlikely that using GPT-4 would increase productivity in the case of Vuo.
For the Vuo editor and engine, the code is object-oriented C++ where you really need an understanding of the surrounding code and relationships between classes to be able to make useful changes. The changes are pretty specific to this codebase, referencing objects and their members. Furthermore, changes need to be made in a maintainable way, meaning that we don’t want similar code duplicated in different parts of the project; it should be factored out into a function and reused. None of the above sound like things that GPT-4 would be able to do well. We already use tools such as Clang to check for the kinds of errors that GPT-4 would (and more).
The code for nodes often doesn’t require as deep an understanding of context, but writing the kind of code that GPT-4 would be able to generate is not where we spend most of our time. We already save a lot of time by identifying open-source libraries that implement things we need rather than us reinventing the wheel. Those libraries usually come with example code that we use as a starting point. We also save time by having automated tests check for common problems with nodes, such as crashing if the input is null. Most of our time is spent making sure the nodes work well with other nodes, and debugging weird problems such as bugs in libraries.
OK, thanks for the response Jaymie, that all makes sense. I was just curious. It does seem like an interesting technology and could be a useful tool in some applications but I have reservations about how it could be used and/or relying on it to heavily.