AI definitions

ℹ️ What is Traditional AI, General Purpose AI, Foundation models, Generative AI?

Artificial intelligence has actually been around since the 1950s (see also Dartmouth Conference, 1956).

It is really difficult to stick and propose a single definition for this group of technologies, partly because these terms, originally stem from computer science, are now applied in a wide variety of settings and used in different ways by other sectors, including policymaking, academia, industry and the media. On the other hand, AI landscape is broad, constantly changing and we must acknowledge and try to understand its distinct forms as they emerge.

The terms below can be subject to multiple interpretations and are often poorly understood or erroneously used. Some of these terms refer to components of AI systems, or related components or subdisciplines of AI. Please feel free to add a new term that you want to be defined, a better definition for an existing one or a useful resource.

Artificial intelligence

Artificial intelligence (AI) can be defined as ‘the use of digital technology to create systems capable of performing tasks commonly thought to require human intelligence. AI is constantly evolving, but generally it:

• involves machines using statistics to find patterns in large amounts of data
• is the ability to perform repetitive tasks with data without the need for constant human guidance” (definition by UK Data Ethics Framework - glossary and methodology)

AI system is a “software that is developed with one or more of certain techniques and approaches, such as “machine learning”, and can, for a given set of human-defined objectives, generate outputs such as content, predictions, recommendations, or decisions influencing the environments they interact with” (definition by EU Artificial Intelligence Act - see below).

https://www.gov.uk/government/publications/generative-ai-framework-for-hmg/generative-ai-framework-for-hmg-html#understanding-generative-ai

Traditional AI

Traditional AI, often called Weak AI or Narrow AI, is a subset of artificial intelligence that focuses on performing preset tasks using predetermined algorithms and rules. These systems have the capability to find rules, learn, make decisions or predictions or generate outputs from data without being explicitly programmed to do so.

Here are some examples of traditional AI: voice assistants like Siri or Alexa (expert systems), recommendation systems on Netflix or Amazon (decision trees), Google’s search algorithm, chatbots, and machine translation systems (NLP).

General Purpose AI (GPAI)

General Purpose AI (GPAI) is defined as an AI system that can be used in and adapted to a wide range of applications for which it was not intentionally and specifically designed.

GPAI systemt “is intended by the provider to perform generally applicable functions such as image and speech recognition, audio and video generation, pattern detection, question answering, translation and others; a general purpose AI system may be used in a plurality of contexts and be integrated in a plurality of other AI systems.” (definition by EU Artificial Intelligence Act - see below)

Definitions of foundation models and GPAI are similar and these terms are often used interchangeably (see below)

-One subset of GPAI systems are systems which can create, generate, or develop images, text, codes and so on. These systems are known as Generative AI (see below).

Generative AI

Generative AI is a subset of GPAI systems that can generate something new: text outputs, images, music and even computer code. Generative AI models are fed vast quantities of existing content to train the models to learn to identify underlying patterns in the data set and, when given a prompt, to generate new data that mirrors the training set /patterns(or outputs based on these patterns). Generative AI relies thus on ‘machine learning’ to understand, predict, and create new content from massive amounts of training data: machine learning provides the training that ‘fuels’ the AI to produce its result.

—> The Key difference between traditional AI and generative AI lies in their capabilities and application. Traditional AI systems are primarily used to analyze data and make predictions (’its smart, but it’s not necessarily responsive’) while generative AI can create new data similar to its training data.