I was just going to ask a similar question recently. Ive been working on a side project involving xgboost and was wondering if ML is still worth learning in 2024.
I recently attended an interesting talk at a local conference. It was from someone that works at a company that makes heating systems. They want to optimize heating given the conditions of the day (building properties, outside temperature, amount of sunshine, humidity, past patterns, etc.). They have certain hard constraints wrt. model size, training/update compute, etc.
Turns out that for their use case a small (weights fit in tens of KiB IIRC) multilayer perceptron works the best.
There is a lot of machine learning out in the world like that, but it doesn't grab the headlines.
I have doubts that a simple adaptive building model-based controller wouldn't be better, and interpretable. I wonder why you'd go with a perceptron... those are so limited.
The foundations of ML aren't changing. The models change, the data pipelines become more sophisticated, but the core skills are still important. Imagine you're trying to predict a binary event. Do you want to predict whether a given instance will be a 0/1 or do you want to predict the probability of each instance being a 1? Why? What do all those evaluation metrics mean? Even if you're using a super advanced AutoML platform backed by LLMs or whatever, you still need to be able to understand the base concepts to build ML apps in the real world.
Turns out that for their use case a small (weights fit in tens of KiB IIRC) multilayer perceptron works the best.
There is a lot of machine learning out in the world like that, but it doesn't grab the headlines.