Machine learning research lays the foundation for artificial intelligence

British-Canadian scientist Geoffrey Hinton and American physicist John Hopfield became Nobel laureates in Physics 2024 thanks to their “discovery and invention that enables machine learning with artificial neural networks.

Mark van der Wilk, a machine learning expert at Oxford University, said an artificial neural network is a mathematical structure inspired by the human brain. The human brain has a network of cells called neurons, which respond to external stimuli such as what the eyes see or what the ears hear by transmitting signals to each other. As we learn, some connections between neurons become stronger while others weaken.

Unlike conventional computers, which operate more like reading a formula, artificial neural networks simulate this process. Biological neurons are replaced by simplex computations called “nodes,” and the stimuli they learn from are replaced by training data. Researchers say that allows the network to learn over time, giving rise to the term machine learning.

Before machines can learn, they need another human characteristic: memory. Hopfield developed the Hopfield network, or associative memory, in the early 1980s. The idea behind the Hopfield network is that when an artificial neural network receives slightly erroneous information, it can check through the model. previously stored to find the most relevant data. This is an important step leading to the birth of AI.

In 1985, Hinton announced the Boltzmann machine, his unique contribution to the field of machine learning. Named after 19th century physicist Ludwig Boltzmann, this design uses an element of randomness. This randomness is why today’s AI imaging tools can produce countless variations from the same suggestion. Hinton also demonstrated that the more layers a neural network has, the more complex its behavior and the easier it is to effectively learn a desired behavior.

Despite the above groundbreaking ideas, many scientists lost interest in machine learning in the 1990s. Machine learning requires extremely powerful computers that can process huge amounts of information. It also takes millions of photos for the algorithm to be able to differentiate between dogs and cats. It wasn’t until the 2010s that a wave of breakthroughs revolutionized everything related to image processing and natural language. From reading medical scans to controlling self-driving cars, from weather forecasting to creating deepfake products, the applications of AI today are too numerous to mention.

Hinton won the Turing Award, an award considered the Nobel in computer science. But some experts say he deserves to win the Nobel Prize in Physics because his scientific research paved the way for AI. French researcher Damien Querlioz points out that these algorithms were initially inspired by physics, by placing the concept of energy in the field of computing.