Microsoft Azure AI: AI's path of becomming a commodity

I recently took the Microsoft Azure AI Fundamentals course and did the certificate - mostly out of interest but admittedly also to have something more to show for my AI work than just the software the software I developed. I found it surprisingly interesting, though, since it in a way opens a different perspective on AI.

I was wondering for quite some time why practically all job openings in AI include Python and experience in training your own AI models as a requirement. I was having fun experimenting with my own little NN backpropagation library years ago and also run quite a number of TensorFlow experiments, but since ChatGPT came out and the really big AI boom is rolling on, I basically stopped with that and happily used the OpenAI services for everything I needed. Don't get me wrong - it's fun and there are many usecases where training models makes perfect sense and is even a requirement because of e.g. privacy issues. But for many use cases you can use off the shelf AI services and connect them into your application. How about an application that uses AI Speech speech to text to record commands, run a content safety check for the request while the AI Search finds relevant information from a database where documents are stored that are parsed by the "AI Document Intelligence", and the AI Language service creates an answer that's shown and the response is again spoken by the AI Speech service? Not a single model needs to be trained for that, saving you loads of time starting that up. And even for things that need training your own model, like numerical predictions based your data, you can use the Azure Machine Learning service.

If a new base version of the model comes out you don't have to retrain your models - you just switch and re-test. And you can tell Microsoft to run the service in Germany or use other German providers like Business GPT or adessoGPT, which likely alleviates many of the privacy concerns, right?

That makes AI more of a commodity, which is both good and bad, as all things. It takes out a bit of the deep magic behind AI, but opens huge perspectives for lightweight AI applications. Interesting #AI times.