> People, even machine learning people, often don't realize the rapidly diminishing returns you get for adding a lot of complexity in your models. Are you me?

> People, even machine learning people, often don't realize the rapidly diminishing returns you get for adding a lot of complexity in your models. Are you me? I was just having this argument at work with someone about using an old (fasttext/word2vec) model vs. the overhead on fine-tuned BERT model for a fairly simple classification problem.