I’ve noticed I was slightly misunderstanding the categorization of AI when I was reading. DEEP LEARNING by Ian Goodfellow, Yoshua Bengio, and Aaron Courville.
First of all, like the drawing above, AI is making a machine do a task that would be otherwise be done by a human. You do some coding to create a program and have the machine execute that. Nothing that special. Just because you wrote a code that calculates 1+1, BAM!, you’ve technically created an AI.
Then there is MACHINE LEARNING. Instead of providing an algorithm to obtain the desired output, you tell the machine which features to consider and have it learn how to map the output from the features you’ve assigned.
Then there is REPRESENTATION LEARNING. When there are SO MANY features to consider, it can easily become computationally expensive. So, what we can do is have the machine learn the features itself as well. In other words, the machine learns the representations(features) to get the desired output.
Finally, there is DEEP LEARNING. When it comes to complex tasks, just a simple REPRESENTATION LEARNING won’t do the work. In DEEP LEARNING, we add more layers, and by doing that the machine will be able to capture more complex and abstract features.
So what I’m trying to say here is that DEEP LEARNING is kind of REPRESENTATION LEARNING, which is also kind of MACHINE LEARNING, which is also kind of AI.
