I’d die on the hill that there is so much more to ML and AI than LLMs. For instance, vector embeddings and text classifiers. So small, and so efficient. I have been able to run hundreds of targeted classifiers as cron jobs sifting through tons of online content - out of a mere 3$/m VM.
@preslavrachev Many people who use spam filters like Spam Assassin actually are using an ML model called naive Bayes classifier. There are many more. ML can be more than just a slop generator and to serve a good purpose.