By saying low level programming I most mean:
- Embedded software - Reverse engineering - OS kernel, driver dev - Complier dev - etc. that is low level in common sense (e.g. Hobbyist APPLE ][ dev)
What kind of impact recent AI tech do you think will have on those fields?
I'm thinking maybe it's going to help a lot with boilerplate code and maybe also help reverse engineering (I think we already have a lot of automation in this field). What others are in your mind?
Most low-level code is detail-oriented, e.g. caring about zero-bug edge cases. Often, there are very few code examples available online. For example, I often must refer to manuals when writing assembly instructions for specific microcontrollers.
I don't know what was in mind for reverse engineering, but typically this involves understanding the flow of a large program and using logical reasoning. So far, these are some of AI's largest weaknesses. Perhaps it could help quickly understand some smaller code segments?
I did try co pilot to generate assembly instructions for a nested interrupt scenario for a particular micro processor that I was well versed with. Code generated wasn't bad. Like any other generated code, one needs to explicitly verify if it's doing the job or it was supposed to.
Co pilot on C: just tried well known algorithms. Task was easier for co pilot here.
Overall, it'll help a programmer work more efficiently with some assistance in hand. And will evolve as we grow towards next generation GPTs
Good amount of time is spent in embedded world to debug chipset specific and hardware specific issues - more so during early days of development. A copilot in this in its current form may not be of great help. Might be in future with enough trained data.
With a little more training, it could be very useful. Right now it's only good for rubber ducking.