The science that built the AI revolution

In a four-part series, we examine the evolution of generative AI. What were the scientific and technological developments that turned the very first, clunky artificial neurons into the astonishingly powerful and versatile large language models that power apps such as ChatGPT?

Part one: What is intelligence? In the middle of the 20th century, the inner workings of the human brain inspired computer scientists to build the first “thinking machines”. But how does human intelligence relate to the artificial kind?

Part two: How do machines learn? Learning is fundamental to artificial intelligence. It’s how computers can recognise speech or identify objects in images. But how can networks of artificial neurons be deployed to find patterns in data, and what is the mathematics that makes it all possible?

Part three: What made AI take off? A decade ago many people working on AI were focused on building algorithms that would allow machines to see and recognise objects. In doing so they hit upon two innovations—big datasets and specialised computer chips—that transformed the entire field of computer science. How did the growth of the world wide web and the design of 3D arcade games create a turning point in the quest to build intelligent machines?

Part four: What made AI models generative? In 2022 the much-anticipated AI revolution finally arrived. With the arrival of ChatGPT and its human-like text, the world finally woke up to the potential of large language models. Remarkable AI-generated images (and also deepfakes) became more pervasive. Suddenly, it seemed that AI models had developed creative skills. Underneath the new “generative” AI models were old algorithms that had been taught some new tricks. And these new models promised to transform…everything.

Listen on: Apple Podcasts | Spotify

On Thursday April 4th, we’re hosting a live event where we’ll answer your questions on AI. If you’re a subscriber, you can submit your questions and find out more at economist.com/aievent.

Podcast transcripts are available upon request at [email protected]. We are committed to improving accessibility even further and are exploring new ways to expand our podcast-transcript offering.