Skip to content

Knowledge that lights the world

What Is Deep Learning? Artificial Neural Networks Explained Simply

What is deep learning and how is it different from machine learning? How neural networks learn, the 2012 turning point and uses from face recognition to chatbots.

What Is Deep Learning? Artificial Neural Networks Explained Simply

Deep learning is an artificial intelligence method that lets computers learn patterns on their own from large amounts of data using many-layered artificial neural networks. In answer to “what does deep learning mean?”, the word “deep” refers to the large number of layers in the network. Today many technologies, from face unlock on your phone to chat assistants, are built on it.

How does a neural network work?

A neural network is made of interconnected artificial “neurons”. Each neuron multiplies the numbers it receives by certain weights, adds them up and passes the result to the next layer. The first layers capture simple features (edges and colours in an image); later layers combine them into more complex concepts (an eye, a face, a cat).

Learning happens during training: the network makes a prediction, the prediction is compared with the correct answer, and the error is spread backwards with a method called backpropagation, adjusting the weights in small steps. Repeated over millions of examples, this lets the network perform tasks for which nobody wrote explicit rules.

What is the difference between deep learning and machine learning?

In classic machine learning, humans often decide which features of the data to look at. In deep learning, the network extracts the important features itself from raw data (pixels, sound waves, text). The price is that it needs far more data and computing power, especially graphics processors (GPUs).

Why did deep learning take off in 2012?

The idea of neural networks goes back to the 1950s, but for decades there was not enough data or computing power. In 2012 AlexNet, developed by a team at the University of Toronto, won the ImageNet image recognition competition by a wide margin and showed the power of deep networks trained on GPUs. The field’s pioneers Geoffrey Hinton, Yann LeCun and Yoshua Bengio received the Turing Award in 2018, and Hinton shared the 2024 Nobel Prize in Physics with John Hopfield.

Where is deep learning used?

  • Images: face recognition, medical image analysis, perception in self-driving cars.
  • Speech: voice assistants and speech-to-text.
  • Language: machine translation and large language models such as ChatGPT, Gemini and Claude.
  • Science: protein structure prediction, weather models and drug discovery.

Frequently asked questions

Are deep learning and AI the same thing?

No. AI is the broadest concept; machine learning is a branch of it, and deep learning is a branch of machine learning.

What do I need to know to learn deep learning?

Basic Python programming plus high-school linear algebra and probability is a good start. You can continue with free libraries such as PyTorch and TensorFlow and online courses.

What are the weaknesses of deep learning?

It needs a lot of data and energy, its decisions are hard to explain (the “black box” problem) and it can carry biases from its training data.

Sources

  1. ACM, 2018 A.M. Turing Award announcement (Bengio, Hinton, LeCun).
  2. Nobel Foundation, 2024 Nobel Prize in Physics (Hopfield and Hinton).
  3. Krizhevsky, Sutskever, Hinton, “ImageNet Classification with Deep Convolutional Neural Networks”, 2012.
Share
Follow YulaHubGet the day's most important stories first.
RSS

Leave a comment

Your email address will not be published. Required fields are marked *