Introduction
Machine Learning in the context of AI
Often times, these two phrases are used interchangeably. Rather, we should see machine learning as one of the tools that is used in the quest to develop artificial intelligence. See more on some of the ideas behind AI here.
Say we want to write a program that recognizes an image chair. Let’s define a chair as “a seat, back, and four legs.”
Not a bad definition, right? What about this one?
Maybe we were a little too specific, how about “something with a seat and a back”?
“Something with a seat?”
You get the idea. Our brains are ridiculously powerful! We’re able to define and classify things as chairs, or cats, or dogs on the go. But how do our brains process these things? We don’t know exactly know how it goes from input to output here.
Alan Turing came up with the analogy that inspires a lot of this thinking. Consider a child with a bunch of “blank sheets,” and a teacher. The teacher doesn’t quite know what goes on in the child’s head, but gives feedback on whether the child did a task correctly or incorrectly. In other words, the teacher rewards good behavior and penalizes bad behavior. In the context of ML and linear regression, we can consider these:
- Child = Model
- Blank Sheets = Parameters (w, b)
- Teacher = Loss Function
So using this idea, we can get started.