Learning reinforcement pioneers win the Turing Award

In the 80s, Andrew Barto and Rich Sutton are considered eccentric devoted to an elegant, but eventually doomed idea – having machines to learn, as humans and animals do, from experience.

Decades onwards, with the technique they are pioneered now increasingly critical to the modern artificial intelligence and programs such as ChatgptBarto and Sutton have been awarded the Turing Award, the highest honor in computer science.

Barto, Professor Emeritus at the University of Massachusetts Amherst, and Sutton, a professor at the University of Albert Experimenting combined with positive or negative feedbackS

“When this work started for me, it was extremely gentle,” recalls Barto with a smile, talking about Zoom from his home at Massachusetts. “It is remarkable that (she) has made some influence and some attention,” Barto adds.

Strengthening training was probably the most famous Used by Google Deepmind in 2016 to build AlphagoA program that learned about yourself how to play the incredibly complex and fine desktop game at the expert level. This demonstration aroused a new interest in the technique that continued to be used in advertising, Optimizing energy use of data centersfinance and Chip designS The approach also has a long history in roboticsWhere it can help machines learn to perform physical tasks through experience and error.

Most recently, reinforcement training is crucial to the production management of large language models (LLM) and the creation of extremely capable chatbot programs. The same method is also used to train AI models imitating human reflectionsand to build more capable AI agentsS

However, Sutton notes that the methods used to lead LLM include people providing goals, not an algorithm who learn purely through their own research. He says that machines can be fully learned in themselves can ultimately be more useful. “The big division is whether (AI) to learn from people or to learn from your own experience,” he says.

“Barto and Sutton’s work has been a Lynch of AI progress over the past few decades” Jeff DeanSenior Vice President on Google, said in a statement published by Computing Machinery Association (ACM) that distributes the Turing Award. “The tools they have developed remain the central pillar of the AI ​​boom and have achieved great achievements.”

Strengthening has a long and plaid history within AI. He was there at the dawn of the field when Alan Turing He assumed that machines could learn through experience and feedback in their famous 1950 document. “Computing machines and intelligence“, Which explores the idea that a machine can one day think as a person. Arthur Samuel, AI pioneer, uses strengthening training to build one of the first machine learning programs, A system capable of playing controlsIn 1955