How important is learning AI/ML now?

Question

It seems like ChatGPT is rapidly taking over the world and similar programs are sure to follow suit. The ability for it to generate usable code on prompting as well as tailor it to your specifications seems like a mandatory skill for programmers. But if that were to start upending jobs, it seems like understanding the nitty gritty behind its implementation would be necessary to stay competitive.However, I was talking with my brother about machine learning since he has been learning and playing with it for a while now, and he said the most important thing isn't the algorithms themselves, but the data they use to train the models. He said the data has to be really well-curated and selected, or the training will end up being basically useless.So is it more important then to become competent at all aspects of data science? I want to not be rendered replaceable due to the AI/ML technology job takeover. Currently I have little experience with either, but I do have about 10 years experience in software development in general.

PaulHoule · Accepted Answer

Your brother has it right.It is very hard to make a language model as good as OpenAI but it is not hard to download a model from huggingface and fine-tune it to solve a particular problem.My favorite kind of ML paper is when someone applies several common models to a random problem they cared about, as opposed to papers where the best in the world compete on problems that have been done to death. I will not get best in the world results but the run of the mill papers help me understand the results that I can get on my own run of the mill problems.

gsatic · Answer

In pure software terms it's quite boring. Worked on speech rec. Most of the time you are just making tools for a giant army to manually clean up/handle better all the edge cases. Game Dev is much more interesting these days.