Machine Learning or Deep Learning?

AI – Artificial Intelligence is the tree, with ML – Machine Learning and Deep Learning being the branches that make up the tree, I’m not saying the tree is only made of ML and DL, but these are a big part of what AI is built upon.

Machine Learning

Machine Learning is all about using known data to predict outcomes based on unknown data that meets certain expectations. This is all about using algorithms to evaluate data, learn from the data and to make informed decisions based on that data.

ML involves building and training a model. ML doesn’t have the ability to learn something new. At least not on it’s own…

Deep Learning

DL is designed to try and mimic the neural networks in a human brain. We might use layers of algorithms to create this artificial neural network that has the ability to learn and improve. DL should be able to make some kind of intelligent decision on it’s own.

Artificial Intelligence

Deep Learning is essentially a subset of Machine Learning and both are subsets of Artificial Intelligence. Generally it is accepted that DL is the more human like AI when compared to ML alone.

ai-dl-ml

Google AlphaGo

The Google AlphaGo was a Deep Learning system designed by Google to play a game called Go. This fast paced board game is based on intuition and intellect. Google matched their system against Go players, the system learned from the games it took part in and improved, eventually went on to beat multiple masters of the game and is now considered to be one of the greatest Go Players of all time.

The reason this made such a stir was that the system continued to learn how to play without ever been taught when to play a certain piece as would be the case in Machine Learning. This system has continued to evolve and back in 2017 entered into a competition in South Korea watched by over 200 million people world wide.

AlpaGo was further enhanced by a new system called AlphaGo Zero, the impressive thing here is that whilst the original system developed it’s mastery of the Go game by playing thousands of Professional and Amateur Go players across the world, the AlphaGo system learned by playing against itself using random starting points, these kind of advancements are built upon the concepts described in Deep Learning.