What is artificial intelligence (AI)?

What is AI?

Artificial intelligence (AI) is a technology that enables computers and machines to simulate human learning, understanding, problem-solving, decision-making, creativity, and autonomy.

AI-powered applications and devices can observe and identify objects; understand and respond to human language; learn from new information and experiences; create detailed recommendations for users and experts; and act independently without the need for human intelligence or intervention (such as self-driving cars).

Machine Learning

A simple way to think about AI is that it's a series of nested and derived concepts that have emerged over more than 70 years.

How are artificial intelligence, machine learning, deep learning, and generative AI related?

Just below AI lies machine learning, which trains algorithms to create models that make predictions and decisions based on data . This includes a wide range of techniques that enable computers to learn from data and make data-driven inferences, even without being explicitly programmed for specific tasks.

There are various types of machine learning methods and algorithms, including linear regression , logical regression , decision trees , random forests , support vector machines (SVMs) , k-nearest neighbors (KNNs), and clustering . Each of these approaches is suited to different types of problems and data.

However, one of the most popular types of machine learning algorithms is called a neural network (or artificial neural network). Neural networks are modeled based on the structure and function of the human brain. A neural network consists of layers of interconnected nodes (similar to neurons) that work together to process and analyze complex data. Neural networks are well-suited to tasks such as identifying complex patterns and relationships within large amounts of data.

The simplest form of machine learning is called supervised learning , where an algorithm is trained using labeled datasets to classify data or accurately predict outcomes. In supervised learning, a human matches each training example with an output label. The goal is for the model to learn the mapping between the inputs and outputs of the training data so that it can predict the labels of new, unknown data.

AI vs Machine Learning (5:49 min)

Deep learning

Deep learning is a subset of machine learning that uses multilayer neural networks, or "deep neural networks," to better mimic the complex decision-making capabilities of the human brain.

Unlike the neural networks used in classical machine learning models, which typically have only one or two hidden layers, deep neural networks contain at least three, and usually hundreds, hidden layers in a single input layer, along with one output layer.

These multiple layers enable unsupervised learning , which means automating the extraction of features from large, unlabeled, unstructured datasets and allowing for independent predictions about what the data represents.

Because deep learning requires no human intervention, it enables extremely large-scale machine learning. It is well-suited for natural language processing (NLP) , computer vision , and other tasks that require the rapid and accurate identification of complex patterns and relationships within massive amounts of data. Deep learning, in several forms, underpins many artificial intelligence (AI) applications in our lives today.

In deep neural networks, multiple layers of nodes can extract meaning and relationships from large amounts of unstructured and unlabeled data.

Deep learning makes the following possible:

• Semi-supervised learning combines supervised and unsupervised learning using both labeled and unlabeled data to train AI models for classification and regression tasks.
  

• Self-supervised learning generates implicit labels from unstructured data rather than relying on a labeled dataset of monitoring signals.
  

• Reinforcement learning learns through trial and error and reward functions, rather than extracting information from hidden patterns.
  

• Transfer learning uses knowledge gained through one task or dataset to improve the performance of a model on another related task or a different dataset.

Generation AI

Generative AI, sometimes also called "generative AI," refers to deep learning models that can create complex original content, such as long texts, high-quality images, realistic videos, and audio, in response to user prompts and requests.

Broadly speaking, generative models encode a simplified representation of the training data, extract data from that representation, and create new data that is similar to, but not identical to, the original data.

Generative models have long been used in the field of statistics to analyze numerical data. However, in the last decade, they have evolved to analyze and generate more complex data types. This evolution has coincided with the emergence of three types of sophisticated deep learning models.

• Variational autoencoders (VAEs), introduced in 2013 , enable models that can generate multiple variations of content in response to prompts and instructions.
  

• This diffusion model, first introduced in 2014, adds "noise" to an image until it is no longer recognizable, and then removes the noise according to prompts to generate the original image.
  

• Transformers (also known as transformer models) are trained on sequence data to generate augmented sequences of content (such as words in a sentence, shapes in an image, frames in a video, or commands in software code). Transformers are at the core of most of the generative AI tools featured in the media today, including ChatGPT, GPT-4, Copilot, BERT, Bard, and Midjourney.

AI Academy

Become an AI expert

Gain the knowledge to prioritize investments in AI that drive business growth. Try the free AI Academy today and lead the future of AI for your company.

View the series

How Generative AI Works

Generative AI generally operates in three phases.

1. Training: Creating a foundational model.

2. Tuning : Adapting a model to a specific application.

3. Generation and evaluation to improve accuracy , and further tuning .
   

study

The starting point for generative AI is the "foundation model." This is a deep learning model that forms the basis for different types of generative AI applications.

Currently, the most common foundational models are Large-Scale Language Models (LLMs) created for text generation applications . There are also foundational models for image, video, audio, or music generation, as well as multimodal foundational models that support multiple types of content.

To create the foundational model, deep learning algorithms are trained on massive amounts of unstructured, unlabeled, and relevant raw data, such as terabytes or petabytes of data, text, and images and videos from the internet. This training creates a neural network consisting of billions of parameters .(Encoded representations of entities, patterns, and relationships within the data) are generated, and content can be autonomously generated in response to prompts. This is the foundational model.