27/06/2017
Know the Difference Between AI, Machine Learning, and Deep Learning:
1. Artificial Intelligence
As a branch of computer science, AI is an area of research aiming to reproduce the various cognitive capacities of human sentience, especially the ability to solve complex problems, in machines. Admittedly, that’s one broad definition of AI, which is a broad and fertile domain itself, open to other scientific and technological disciplines.
AI may refer to NPCs (non-player characters) in video games, image recognition systems, voice and speech recognition platforms, autonomous vehicles, predictive algorithms and other specialized computer programs.
All these forms of AI have one thing in common: they are based on pre-defined input, in other words, programmed beforehand to carry out a specific task. That leads us to the next level, machines that learn by themselves.
2. Machine Learning
A subset of AI, Machine Learning focuses on learning abilities, or how to make machines learn on their own. Without the need to hand-code instructions, ML systems get access to large datasets, apply their knowledge, train and learn from mistakes to complete a specific task.
For example, IBM’s Deep Blue–which beat chess master Garry Kasparov in 1997–is not strictly a Machine Learning system because it wasn’t able to cross-reference past moves and matches.
However, the Google AlphaGo that beat world champion Lee Sedol at the game of Go is a machine learning platform that recalled hundreds of previous plays to inform its tactics.
ML systems sift through data, learn patterns and predict outcomes, that why Machine Learning tools are at the top of interests for data-driven businesses.
3. Deep Learning
As a subfield of Machine Learning, and a sub-subset of AI, Deep Learning is automatic learning technology based on deep neural networks. When you know the difference between AI and terms that further define it, you are practicing the very concept applied to these systems.
A DL algorithm is made of layers of artificial nodes or “neurons”, forming sort of a virtual computer where each layer performs simple calculations that serve as input to the following layer. That translates to huge gains in time and efficiency of the whole system. Like with machine learning, these systems are also free of catastrophic forgetting, in that they are able to recall data from past computations and apply them to present solutions.
Expensive and requiring huge storage and processing resources to train, DL systems are being developed by major companies, such as Amazon, Facebook, Google, IBM, and Microsoft, and will almost certainly be given the power to make influential decisions in the near future.
