Artificial Intelligence, Machine Learning or Deep Learning; What’s the difference?

We use the terms AI, ML, Deep Learning and many others more and more these days, but what does each mean, and how are they different? In the first in a series of articles, we will look at some of these fundamentals.

First, and most importantly, AI, ML and DL are all related. Deep learning is a subset of Machine Learning, which is itself a branch of Artificial Intelligence. Think of artificial intelligence as the sky, machine learning as a cloud and deep learning as a raindrop in the cloud.

Artificial intelligence is the engineering of intelligent machines, systems and programmes. They may be able to sense, reason and adapt, mimicking human intelligence. How they do thi s isn’t specified, hence the more details subsets.

Computer programmes have, by and large, always been a set of instructions: if x happens, do y. However, in 1959 Arthur Samuel started to discuss if, rather than being a strict set of rules, computers could learn behaviour. Machine learning is enabled by statistical models and algorithms rather than following specific instructions. A great example is image recognition; over time, a computer can learn to identify particular images. Say we want to identify cats in pictures (I mean, how doesn’t). If we were going to programme a computer to do this, we would give it a specific set of instructions. For example, we may say a cat is about 45cm long, has fur, a tail, claws, etc. In machine learning, we don’t do this; we show the machine lots of pictures of cats and let it figure it out.

Deep learning is a discipline of machine learning. I won’t go into it too deeply (pun intended) here, but essentially it is (often) a neural network that is capable of unsupervised learning from unstructured data. An example of deep learning is automictically captioning images. Say a machine is shown a picture of a child jumping on a trampoline. It would identify the trampoline, the child, and the activity producing a caption “child jumping on a trampoline” from everything it knows.

We have only touched the surface here, but over the coming weeks, we will delve into more detail around how ML works, how it is applied, along with the pros and cons of modern artificial intelligence.

Other posts: