The following examples come from the past 50 years of machine learning research. They will help you to understand the large amount of work in the subject of AI and will inspire your own work in AI.

​

Choose one or two of these examples to explore further using the links below.

​​

Autocorrect

Autocorrect, a feature we take for granted in modern digital communication, employs AI algorithms to predict and correct spelling or typing errors in real time. Autocorrect started as an algorithm that used simple rule-based systems, but it has evolved significantly through AI techniques like natural language processing and machine learning. These algorithms analyse massive amounts of text data to learn patterns, context, and common mistakes. This allows the program to suggest or automatically replace misspelt words with the most probable alternatives. Modern autocorrect includes continuous learning, using user interactions and corrections that allows it to adapt and become more accurate over time.

 

Tic tac toe system by Donald Michie 

In 1960 an AI researcher named Donald Michie developed a technique for solving how to play tic-tac-toe, also called noughts and crosses.

You can play tictac toe online here

He realised that he could represent each state of the game with a small box, and he used matchboxes to do this. Each box contained a number of beads. All the boxes together, even though it was a physical representation instead of a computer program, can be seen as a neural network. Because of the clever way that Donald designed the workings of the boxes, the whole system “learnt” how to solve tic-tac-toe. 

Positions in tic-tac-toe were represented by different coloured beads. A human would play against the system by choosing one coloured bead from the box that shows the current board layout. If the human won the game, all the beads chosen would be taken out of those boxes. If the system won, three beads the same colour as the ones used would be added to the right boxes. This meant that after many games, there would be more of some beads in the boxes than others, so playing the game would mean that the beads that are better at winning the game would get picked more often. This shows that even a simple, mechanical system can “learn” and change over time.

• Find out more about Donald's sytem here: https://opendatascience.com/menace-donald-michie-tic-tac-toe-machine-learning/

• Have a go at using a web simulation of the system here: https://www.mscroggs.co.uk/menace/

• Learn more about the system with the video below