I still feel a bit strange about calling someone an “AI engineer” and I think there are a few reasons. 1. AI is poorly defined. This is a fundamental problem b

I still feel a bit strange about calling someone an “AI engineer” and I think there are a few reasons. 1. AI is poorly defined. This is a fundamental problem behind almost every conversation on the “topic”. Depending on the context, AI can mean anything from a decision tree to deep neural networks to science fiction. 2. Engineering implies a deeper level of understanding. If you want to engineer a system, you need a deeper understanding of how each of the components work. Using a tool does not make one a tool engineer. It makes them a 21st century blacksmith. Does calling an LLM API make me an AI engineer? If so, calling a weather API makes me a weatherman. 3. This is a relatively new area, and make no mistake, it is a new area. 2 years ago most of the tooling around ML work meant you had to get your hands dirty. To use BERT, you pretty much needed to learn about tokenization and attention masks and CUDA. Not so much with GPT3. So it seems premature to even circumscribe at the moment. What I am not saying: I don’t think one needs to take a class in linear algebra to work with this stuff. I also never believed in calculus for CS students, which may be a minority opinion too.