Learning learning pioneers win the Torring Prize

In the 1980s, Andrew Barto and Rich Soton They considered the eccentric worshipers for an elegant but convicted notes in the end – learn stone machines, as humans and animals do, from experience.

After decades, with this technology, which were now increasingly severity of the conversation artificial intelligence And programs like ChatgptBarto and Soton won the Torring Award, the highest honor in the field of computer science.

Barto, Fakhri Professor at the University of Massachusetts Amheres, Soton, Professor at the University Experience with positive or negative reactions.

“When this work began for me, it was very non -modern,” said Barto with a smile. “It was striking that (he has achieved) some influence and some attention,” says Barto.

Perhaps learning to reinforce it was famous Google DeepMind is used in 2016 to build AlphaoIt is a program for himself how to play a complex and hidden plate game that starts at the level of experts. This demonstration has sparked new interest in the technology, which continued to announce, Improving the use of data center energyAnd financing and Frame design. The approach also has a long history in RobotsWhere machines can help learn to perform physical tasks through experience and error.

Recently, reinforcement learning has been very important to direct the output of the LLMS models (LLMS) and the production of unusually capable Chatbot programs. The same method is also used to train artificial intelligence models on Tradition of human thinkingConstructive Artificial intelligence agents are more capable.

However, Sutton notes that the methods used to guide LLMS include humans who provide goals instead of an algorithm that you learn purely by exploring them. He says that having machines that are fully learned on their own may eventually be more useful. He says: “The great division is whether (artificial intelligence) learns from people or whether he learns from his own experience.”

Barto and Soton’s work has been “a work of progress in artificial intelligence over the past few decades,” Jeff DeanGoogle’s first vice president said in a statement issued Computing Machines Association (ACM) that receives the Torring Prize. “The tools that they have developed remain a central pillar of artificial intelligence and make great progress.”

Reinforcement has a long and volatile history within artificial intelligence. There was at the dawn of the field, when Alan Torring I suggest that machines can learn through experience and comments in its famous paper in 1950.Computing and intelligence machines“Who is studying the idea that the machine may one day think as a human being. Arthur Samuel, the pioneer of Amnesty International, used to learn to reinforce to build one of the first automated learning programs, The system is able to play the game1955.