How do you use AI for algorithms implementations?

Hi you all :)

So a bit about myself, I am a computer science researcher, and as a part of my job I write algorithms and need to implement them. Lately, most I do is use the great methods this community has given me and implement the experiments pipeline using AI, but I have an interesting challenge I think you guys can help me with.

As I said, a lot of the work revolves around implementing algorithms, mostly in python. The thing is with my kind of algorithms (and well, lots of the well used algorithms right now), is that just by looking at the input and output I can't easily verify that the algorithm was implemented correctly.

When using AI, it becomes even more of a challenge, as when I implemented things on my own, I understood the code much better, but now Claude can write lots of messy code that understanding it will waste all the time I saved by writing the code with Claude (and besides, I listen to what Jake said, AI is the new level of abstruction, as I am not interested in the C code behind the python implementations, ideally I shouldn't need to understand the python code that the AI from my markdowns).

In other words what I try to ask is, given that the input and outputs are not enough to verify that the implementation of your value is correct, how do you check that you got what you wanted?

Thanks so much in advance, happy to be here :)

0 comments