A short history of AI in 10 landmark moments

From the first digital computers built in the 1940s to the multitudinous uses of today’s machine learning tools, artificial intelligence (AI) has come a long way in a relatively short space of time.

Until recent years, progress was slow compared to today’s standards, with the computational power of AI doubling roughly every 20 months. Fast-forward to 2024 and it is racing ahead, getting twice as powerful every six months.

Here is a brief look at some of the significant developments on that timeline – and what might come next.

This was one of the first examples of machine learning – a robotic mouse that was able to navigate its way through a maze, ‘learning’ as it did so. Developed by Claude Shannon, a US mathematician and researcher, Theseus was a tiny wooden device with metal whiskers on its nose and a magnet in its body. Under the maze, a series of telephone relay circuits and an electromagnet mounted on a motor helped Theseus find its way around, recording exits and walls as it went. As MIT Technology Review reports, Shannon said the machine was “capable of solving [by trial and error], remembering the solution, and also of forgetting it in case the situation changes and the solution is no longer applicable”.

Officially known as the Dartmouth Summer Research Project on Artificial Intelligence, this convention held at US university Dartmouth College is widely considered to be the birthplace of AI as a field. A small team of scientists – including Claude Shannon – coined the term artificial intelligence for the event and set the path for future thinking and research on the technology.

A significant step in machine learning, Perceptron is regarded as the first artificial neural network, which is a program that makes decisions in a manner similar to the human brain. Developed by US psychologist Frank Rosenblatt, the model taught itself to distinguish between punch cards marked on the left and right. Rosenblatt described it as “the first machine which is capable of having an original idea”.

The Adaptive Linear Element, or ADALINE, was a simple, or single-layer, artificial neural network developed by Bernard Widrow, a professor at Standford University in the US and his student at the time, Ted Hoff. An adaptive system for pattern recognition, it laid the foundation for future advances in neural networks and machine learning.

Between these rough dates, the world experienced lulls in AI funding, research and development often referred to as AI winters. There were still some significant developments, however, including a program called TD-Gammon that in 1992 learned to play the board game backgammon at a level that was just below the top players of the time.

IBM’s Deep Blue went one further than TD-Gammon and all other machines that had existed before it – becoming the first computer system to defeat a reigning world chess champion in a standard tournament match. Deep Blue’s underlying technology advanced the ability of supercomputers to tackle the complex calculations needed to perform tasks including uncovering patterns in databases.

The introduction of AlexNet, a deep-learning neural network with many layers, marked a breakthrough in image recognition. It could recognize images of objects such as dogs and cars at a level very close to humans.

While many major tech companies are involved in the evolution of AI, it was the release of then-little-known OpenAI’s Generative Pre-trained Transformer 2 (or GPT-2) that showcased the power of natural language processing. Able to predict the next item in a sequence, it could perform tasks such as summarizing and translating text. “Our model is capable of generating samples from a variety of prompts that feel close to human quality and show coherence over a page or more of text,” OpenAI said in a blog at the time.

The release of OpenAI’s GPT-3 in 2020 brought a model that was able to produce text that is often indistinguishable from that written by humans. The chatbot ChatGPT, released in late 2022, was built on this large language model – and introduced generative AI proper to the wider world.

ChatGPT’s release sparked a new phase of rapid development, and generative AI quickly began to transform every aspect of business and our lives. Alongside, challenges with bias and hallucination brought calls for generative AI to be implemented and governed responsibly. Initiatives such as the World Economic Forum’s AI Governance Alliance are working to ensure AI is inclusive, ethical and sustainable as it drives change across industries.

But alongside ensuring the benefits of AI reach everyone, what’s next for the technology?

Experts say we should expect interactive AI – bots that can instruct other software to carry out tasks for you – AI making new scientific discoveries, and models that understand the physical world, remember, reason and plan.