What is Machine Learning? 🔗
Machine learning, is just another way of writing programs. A program can be represented by an input, a set of rules or statements (the program) and an outcome.
The traditional way of writings software is to recollect the behavior of the system from the users, and write the rules that matches this behavior.
For most problems this works well, but for some other it’s really hard. i.e getting the intent of a sentence. Why? Cause you need to tell every small step to the computer, and we don’t know exactly which steps are.
Machine learning, is making this process up-side-down, giving examples of matched input and output, and letting the computer figure out the rules of the program.
Caveats 🔗
A compromise is that, now the behaviour of the program is driven by the data, we cannot understand the program. the model is not describable
Also any kind of incorrect data / bias will imply errors in the program
Reference: 🔗
Fast AI book: https://github.com/fastai/fastbook/blob/master/01_intro.ipynb