How does AI work and is it always right?

EVERYBODY seems to be talking about how AI is going to change the world – but what on Earth is it?

The technology has rapidly advanced over the past few months, leading many to scrub up their knowledge of it.

While Siri gives users the impression it is artificially intelligent, Siri is a digital voice assistantCredit: Alamy

What is AI?

AI stands for artificial intelligence and it lets computers to act in a more ‘human’ way.

This includes being able to hold conversations – both casual and intellectually challenging ones – with humans.

Computer scientist John McCarthy, one of the founders of AI as a concept, put it like this: “It is the science and engineering of making intelligent machines, especially intelligent computer programs.

“It is related to the similar task of using computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable.”

Google launches AI chatbot Bard that contains 'the breadth of world's knowledge'

Five jobs at risk of vanishing with 'humans replaced by AI bots'

How does AI work?

AI can analyse a vast amount of data at speeds that are impossible for human minds.

It can be trained, as well as learn.

The basis of AI is machine learning, which relies on models using maths, statistics and classification methods.

Developers will teach the computer to understand classification methods, which means the AI can spot patterns and make groups of particular things.

Most read in Tech

A machine with in-built AI can learn from its interactions with human users, individual data sets and the internet.

All it needs to do first is it understand the learning process, and developers will program this understanding into the machine.

Is AI always right?

Today, AI can help doctors diagnose patients, pilots fly commercial aircrafts and city planners predict traffic.

While it can do these things well, this doesn’t mean it is always right.

An AI bot is typically smarter than the average person, but it has been developed by humans – meaning it can be susceptible to human error and bias.

AI bias can see a computer make decisions based on training data which can reflect human biases like social inequalities.

In 2018, Amazon stopped using a hiring algorithm when it discovered the AI favoured applications based on words like “executed” or “captured” – which were more commonly found on the CVs of men.

What are the four types of AI?

Reactive AI – the most basic form of AI which lacks the ability to form memories or use past experiences to inform current decisions.
Limited Memory AI – a type of AI that is commonly found in self-driving cars and allows machines to look into the past to understand lane markings, for example.
Theory of Mind AI – the AI understands that people, creatures and objects in the world can have thoughts and emotions that the machine will have to change its behaviour to suit.
Self-aware AI – the final step of AI, where developers understand consciousness to the point where they can program a machine to have it.