deep learning (DL)

basicwidespread

Deep learning is a method of machine learning in which a multi-layered artificial neural network learns on its own to recognize patterns from a vast number of examples.

In plain terms

Imagine that, instead of explaining to a child what a dog is with rules, you show them lots of pictures of dogs. After a while, they sense on their own what makes a dog a dog. Deep learning does something similar with machines, except the machine needs far more examples. The programmer does not write the rules; the system derives them from the examples.

The word 'deep' does not refer to the depth of understanding but to the fact that the network consists of many stacked layers. It is like an assembly line: the first station detects only lines and spots in an image, the next recognizes shapes, and the later ones identify faces, cars, or letters.

This technique underlies almost everything we call artificial intelligence today: chatbots, image generators, translation programs, speech recognition, and AI agents. It differs from previous methods in that it can learn from raw data (images, audio, text) and does not need to be told in advance what to focus on.

How it works

The network consists of small computing units, artificial "neurons," arranged in layers. Numbers define the strength of the connections between them, called weights. A large model can contain billions of such numbers.
During training, the system receives an example (say, an image) and guesses. Initially, the guess is practically random because the weights start as random values.
The guess is compared with the correct answer, and the error magnitude is calculated. The procedure called backpropagation calculates, moving backward from the output, which weight needs to be adjusted in which direction. Then every weight is changed slightly so that the network makes smaller errors next time.
This cycle repeats millions or billions of times on massive datasets. The calculations are performed by powerful graphics processing units (GPUs) or other chips designed for this purpose, lasting days or even months.
The weights of a trained model generally do not change during use. The model answers new, previously unseen data using the weights as set. This is what happens when you ask a chatbot something.

An everyday example

When your phone unlocks with facial recognition, or you search for 'beach' in the gallery and your vacation photos appear, deep learning is working in the background. The same happens when a translation program translates whole sentences meaningfully, or when a chatbot answers your question.

Who is affected

everyoneYour phone's camera, translator, search engine, and chatbot are all built on this, so you use it daily even if you are unaware of it.

employeeAI tools appearing in workplaces are almost all based on deep learning models, so understanding this helps you know what to trust in them and what not to. → Work & HR

patient, doctorPrograms analyzing X-ray, CT, and dermatological images use deep learning to search for suspicious lesions, but the decision remains with the doctor. → Health

developerIt is the foundational technology of modern AI systems: everything from large language models to coding agents is built on it. → Code

A common misconception

MythA deep learning system thinks and understands like the human brain, only faster.

In realityA neural network only loosely resembles the brain: in reality, it searches for statistical patterns through massive amounts of calculation. Whether this can be called understanding is debated among researchers.

Limits and risks

It requires vast amounts of data, computing power, and energy, so only a few well-funded companies can train the largest models.
It is hard to say exactly why it gave a certain answer: its operation is often a "black box", which is a serious problem in healthcare, banking, or legal decisions.
It learns what is in the data, meaning it also adopts the errors and biases contained within.
It can err confidently: in unusual situations differing from the training data, it may be unreliable and often fails to signal this.

Similar, but not the same

machine learning – Machine learning is the broader category that includes simpler methods; deep learning is its branch based on multi-layered neural networks.
artificial intelligence – Artificial intelligence is the umbrella term for the entire field, while deep learning is just one of its techniques – currently the most successful.
large language model – A large language model is an application of deep learning specialized in text; deep learning can also be used to analyze images, audio, or protein structures.

Related concepts

Broader

artificial intelligence, machine learning

Narrower

large language model, transformer, convolutional neural network

On the edge

neural network, deep neural network, generative AI, training data, backpropagation, GPU, parameter

A brief history

The idea of neural networks dates back to the 1940s, but the breakthrough came in 2012 when the deep network named AlexNet won the ImageNet image recognition competition by a large margin. The field's three pioneers, Geoffrey Hinton, Yann LeCun, and Yoshua Bengio, received the Turing Award in 2018, and Hinton shared the Nobel Prize in Physics with John Hopfield in 2024.

Where things stand

At the end of 2026, deep learning is the dominant technology in artificial intelligence: chatbots, image and video generators, and agents that carry out tasks autonomously are all built on it. The main question is no longer whether it works, but how much energy it consumes, how reliable it is and how transparent it is.

Status: 2026 November

In the news

Appears in 1 Agent Age stories; the curve shows weekly occurrence in the news magazine.

82% of Indian students fear AI speeds up work at the cost of deep learningOct 19, 2026 · Education & Family

Also known as

deep learning · deep learning · deep learning · DL · deep learning

Sources

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.