what is machine learning?
- my example with linear regression
- decision tree example (show an example of a tree, then explain how the algorithm can learn it)
- most importantly, it requires you to get the data in the format that you have "X" and "y"

what is it to be used for?
- should money be paid to a client (insurance)?
  -> they cannot have an AI say "No" to people (the result has to be explainable)
- automatically detect damage on cars (computer vision at a customer)


warnings:
- using historic data is wrong when you have a new "rule"
- I don't recommend to do machine learning when you still have to collect the data
  - chatbot! you can use customer interactions, it's useful data... can be used in analytical way, but not through machine learning
- when someone says they have "a lot of useful data" -> run!


tips:
- start with a new process that does not rely yet on rules
- in big companies: spend time building a small prediction framework specialized for your core data
- Use AI stories, build towards it, but don't forget the "easy gains" along the way
- sometimes just be excited about automation! that's usually the end goal
- go make a little open source project using machine learning!

tricks:
- machine learning allows people to do a simple task of training without having to add code

when does machine learning help?

requirements when to use machine learning:
- it is very difficult to code the rules
- it too time consuming to manually evaluate all cases