Artificial intelligence vs. machine learning vs. deep learning

Are you a bit confused by all the AI-related terminology? These days, we’re constantly reading about artificial intelligence in the newspapers, on LinkedIn, on social media and even in everyday conversations. However, we probably don’t know exactly what it involves or how it differs from other related concepts that sound rather similar.

In this article, we aim to shed some light on all this jumble of terms and explain the differences between artificial intelligence, machine learning, deep learning and generative AI.

Artificial intelligence (AI)

Artificial intelligence is a field of computer science that focuses on creating systems or programmes capable of performing tasks that normally require human intelligence. These artificial intelligence systems share certain capabilities with humans, such as learning, reasoning, perception, pattern recognition, problem-solving, amongst others. The aim of AI is therefore to simulate human thought processes so that machines can carry out tasks autonomously and make decisions.

After reading that AI aims to simulate, and even surpass, human capabilities, we may feel a certain sense of concern and imagine a dystopian world in which technology ends up subjugating humankind. Nothing could be further from the truth. Artificial intelligence is nothing more than a human invention, controlled by humans, devoid of any real cognitive will of its own and limited to the training datasets it is provided with.

Machine learning

Machine learning is a sub-discipline of artificial intelligence that focuses on the development of algorithms and models which enable computers to learn patterns and make predictions based on data without being explicitly programmed to do so. Rather than following specific instructions, machine learning algorithms use data to learn and improve as they gain experience.

This is the process that engineers and programmers follow to implement machine learning in an artificial intelligence system:

• Data collection: Firstly, we collect data relevant to the problem we wish to solve, ranging from images and text to numerical data.
• Data preparation: We then clean and organise the data so that the computer can understand it better. To do this, we remove unnecessary data, correct errors and format the data appropriately.
• Choosing an algorithm: Next, we choose a machine learning algorithm that is suited to the problem we want to solve and the data we have. There are different types of algorithms, such as regression, classification and clustering algorithms.
• Model training: Now, we feed our data into the algorithm so that it can learn patterns and relationships. During this phase, the algorithm adjusts its parameters to minimise the error between the predictions and the actual results.
• Model evaluation: Once the model has been trained sufficiently, it needs to be tested on a dataset it has not seen before to check that it generalises well and performs with the same accuracy.
• Adjustment and optimisation: If the model does not perform satisfactorily, we adjust the algorithm’s hyperparameters or try different data pre-processing techniques to improve its performance.
• Deployment and use: Finally, once we are satisfied with the model’s performance, it is deployed into production so that it can make predictions or take decisions in real time based on the new data it receives.

Artificial intelligence vs. machine learning

Here is a summary of the differences between artificial intelligence and machine learning, as we have explained above:

Artificial intelligence:

• A broad field of computer science that aims to develop systems capable of performing tasks that normally require human intelligence.
• It encompasses a range of approaches and techniques, such as machine learning, natural language processing and computer vision, amongst others.
• It aims to simulate human intelligence in general.

Machine learning:

• A subfield of artificial intelligence that focuses on the development of algorithms and models capable of learning from data and improving their performance with experience.
• Machine learning algorithms enable computers to recognise patterns and make decisions without being explicitly programmed to do so.
• It is based on training models using data to make predictions or take decisions.

Deep learning

Deep learning is a branch of machine learning based on multi-layer artificial neural networks (also known as deep neural networks) to model and extract high-level representations of data. These networks are made up of several layers of nodes, each of which performs mathematical operations on the input data to produce outputs. Deep learning has proven to be highly effective in tasks such as computer vision, natural language processing and speech recognition, amongst others.

Machine learning versus deep learning

One of the main differences between deep learning and conventional machine learning techniques lies in its ability to handle highly dimensional and complex data. For its part, it is often difficult to handle large volumes of unstructured data or highly complex ones using conventional machine learning algorithms. As a more advanced AI technique, deep learning can process this information more efficiently and effectively.

Unlike machine learning, deep learning – setting aside algorithms and training with datasets –, uses deep neural networks to learn from these large amounts of data automatically. These neural networks are made up of multiple layers of artificial neurons, through which they learn to recognise complex features and perform tasks autonomously.

In practical terms, Machine learning and deep learning are applied in different scenarios. Whilst machine learning works best with structured data, deep learning is more suited to unstructured data.

An example of a deep learning application

Imagine we want to develop a system to detect and diagnose diseases using X-rays or MRI scans. In this case, deep learning would be more suitable due to the complexity and the large volume of unstructured data present in X-rays. We could use a deep convolutional neural network (CNN) architecture to enable the artificial intelligence system to automatically learn to identify relevant features and patterns.

What is generative AI?

It’s highly likely that, amidst all the hype surrounding artificial intelligence over the past couple of years, you’ve heard of applications such as ChatGPT or Midjourney. These are examples of systems that fall under the umbrella of what is known as generative AI. Generative artificial intelligence is a field of artificial intelligence that focuses on the creation of systems capable of generating new and original content, such as images, music, text or even videos, which are indistinguishable (or almost indistinguishable) from those created by humans. In fact, the image at the top of this article was created using generative AI.

These systems use generative deep learning models, which are capable of learning the characteristics and patterns of a given dataset and then generating new examples that resemble that original data. A common example of a generative model is the generative adversarial network (GAN), which consists of two neural networks that compete against each other: a generator, which produces fake samples, and a discriminator, which attempts to distinguish between the fake and real samples.