Is it still worth learning how to work with MIC architecture for machine learning ?
I'm new to AI and using hardware to accelerate machine learning algorithms. I've seen there is quite a big race between several technologies for machine learning such as using GPU, CPUs like Xeon and Xeon Phi , and also FPGAs. I'm interested in learning MIC architecture with Xeon Phi products to implement it in machine learning acceleration. But it seems even Intel has given up on Intel Xeon Phi product line(you can read this news page : https://www.nextplatform.com/2018/07/27/end-of-the-line-for-xeon-phi-its-all-xeon-from-here/). Hence I'm skeptical about investing time on it to try using this technology if it has ended. Could someone explain is it still worth taking time to learn the MIC architecture for machine learning ?
Sorry if the question seems trying to predict the future, but I'm a novice to the field and any advice seems valuable to me.