Here's my cynical take on the state of the art for 99% of AI projects: 1. take some training data. 2. Do some statistical analysis on it and find relevant varia

Here's my cynical take on the state of the art for 99% of AI projects: 1. take some training data. 2. Do some statistical analysis on it and find relevant variables for the variable you want to predict. 3. Build a model to predict the variable. 4. Put it into production. 5. Run model and get predictions. 6. Notice predictions are becoming less reliable 7. Retrain the model on new data. 8. GOTO Step 5.