Mind Machines: how does AI learn?

By copying the behaviour of the brain’s own neurons, it was possible to develop the first ‘neural network’. Neural networks are computer programmes which self-improve and get better over time. During the data explosion of the 2000s, they evolved to become the machine learning and deep learning that we know today where an AI is able to learn, respond and improve by itself.

"Technology that is able to teach itself can endlessly surprise us with its apparent creativity and its ability to see, hear and move."

Ada Lovelace and the first algorithm

Augusta Ada King, Countess of Lovelace was an English mathematician and writer, chiefly known for her work on Charles Babbage's proposed mechanical general-purpose computer, the Analytical Engine.

In 1843, Ada Lovelace translates 'Sketch of the Analytical Engine', invented by Charles Babbage. In the notes to the translation, she develops what is considered the first algorithm to make Babbage's engine compute a full Bernoulli sequence - effectively making her the first computer programmer.

What is a Bernoulli sequence?
The Bernoulli process is a finite or infinite sequence of binary random variables - either 0 or 1.

Charles Babbage and the first steam-powered computer

Charles Babbage was an English computer programmer, mathematician, philosopher, inventor and mechanical engineer. Babbage originated the concept of a digital programmable computer.

The Analytical Engine is a computing machine designed by mathematician Charles Babbage (1791 – 1871). Now recognised as the first steam-powered computer, it could calculate any mathematical formula through its central processing unit and memory function. The machine was only partly built during Babbage’s lifetime, but Ada Lovelace’s writings contained extensive descriptions. In 1843, Lovelace published detailed information relating to the engine, including algebra, mathematics, logic and even philosophy.

The impact of Alan Turing

Alan Turing was an English mathematician, computer scientist, logician and cryptanalyst. He was highly influencial in the development of theoretical computer science and formalised the concepts of algorithms and computation in the Turing machine.

During the Second World War, he was part of the Government Code and Cypher School at Bletchley Park, Britain's codebreaking centre. He was part of the team that was responsible for intercepting German ciphers and played a pivotal role in cracking the messages that enabled the Allies to defeat the Nazis and in doing so, helped win the war.

Cracking the code to win the war

During World War II, the Germans used Enigma’s encryption to protect their communications. Every time a letter was pressed on an Enigma, the machine changed it to a different letter. Enigma messages could only be deciphered using a key, similar to a password.

Keys changed from message to message, and regular improvements to procedures made Enigma messages even harder to decipher. Towards the end of the war, each message had 103,325,660,891,587,134,000,000 (103 sextillion) possible encryptions.

SoFi - The Soft Robotic Fish, MIT

How can technology improve our understanding and access to nature? MIT Computer Science & Artificial Intelligence Lab (CSAIL) have been developing a new kind of robot destined to live underwater.

Using a miniaturized acoustic communication module, a diver can direct the fish by sending commands such as speed, turning angle, and dynamic vertical diving.

Robot puppy, aibo, Sony

Sony’s 2018 robot puppy, aibo, uses its database of memories and experiences to develop its own personality.

Sold from 1999 – 2006, Sony’s AIBO was the world’s first entertainment robot for domestic use. The emergence of AIBO prompted the idea that robots were no longer simply for labour and could also be lovable creatures kept at home.

In 2018, AIBO returned as aibo, with cloud-connected AI capable of evolving through a collective intelligence linking individual robotic dogs. aibo begun its journey in homes and is now also used in care homes and hospitals.

Waterfall of Meaning, Google PAIR

Google PAIR’s project Waterfall of Meaning is a poetic glimpse into the interior of an AI, showing how a machine absorbs human associations between words.

Circuit Training, Mario Klingemann

Artist Mario Klingemann’s piece Circuit Training invites visitors to take part in teaching a neural network to create a piece of art.

Visitors will first help create the data set by allowing the AI to capture their image, then select from the visuals produced by the network, to teach it what they find interesting. The machine is constantly learning from this human interaction to create an evolving piece of live art.

Myriad (Tulips), Anna Ridler

Artist Anna Ridler looks at the politics and process of using large datasets to produce a piece of art. Inspired by ‘tulip-mania’ - the financial craze for tulip bulbs that swept across the Netherlands in the 1630s, she took 10,000 photographs of tulips and categorised them by hand, revealing the human aspect that sits behind machine learning.