The Complexity of ML Models

When we're thinking about developing ML models, then the way that we develop them is, you know, we really sanitize the data. Cause it's not just missing features, data types for the features. And when models don't have that basic sensitization or kind of like basic guardrails or to protect them from something like a missing feature can lead to the model having these sort of serious failures. We need to read your own diagram of that 2015 paper from Google about where the time investment is going into ML because it's so much more complex now.