In the 1980s, Andrew Barto and Ricky Sutton Being considered a eccentric devotee, it is an elegant but ultimately doomed idea – like humans and animals, learn from experience, just like humans and animals.
For decades, they have relied on their technology that is increasingly important to the modern world now AI and similar programs chatgptBarto and Sutton were awarded the Turing Award, the highest honor in the field of computer science.
Barto, professor emeritus at the University of Massachusetts Amherst and Sutton, professor at the University of Alberta, pioneered a technology called reinforcement learning that involves coaxing computers to perform tasks through a computer through a computer. Experiment combined with positive or negative feedback.
“When the work started for me, it was very out of place,” Barto recalled smiling, speaking from his home in Massachusetts. “This is really amazing [it has] Some influence and some attention were achieved,” Barto added.
Reinforcement learning is probably the most famous Alphago used by Google DeepMind in 2016The plan for yourself to learn how to play incredibly complex and subtle board games. This demonstration sparked new interest in the technology, which has been used in advertising, Optimize energy use in data centersfinance and Chip design. This method has a long history in history Roboticsit can help machines learn to perform physical tasks through trial and error.
Recently, reinforcement learning has been crucial to direct the output of large language models (LLMs) and to produce powerful chatbot programs. The same method is also used to train AI models Imitating human reasoningand built More capable AI agents.
However, Sutton points out that the method used to guide LLM involves humans providing goals rather than exploring purely algorithmic learning through its own exploration. He said that letting machines learn completely on their own may end up being more fruitful. “The biggest department is whether it is [AI is] Learn from people or learn from your own experience. ” he said.
Barto and Sutton’s “work has been an advancement in AI for the past few decades),” Jeff DeanGoogle’s senior vice president, Computer Association (ACM) presented the Turing Award. “The tools they develop remain central pillars of the AI boom and have made significant progress.”
There is a long history in AI. It was in the dawn of the fields Alan Turing It is suggested that the machine can learn from the experience and feedback from his famous 1950 papers”Computers and intelligence“, it studies the idea that machines may one day think like humans. AI pioneer Arthur Samuel uses reinforcement learning to build one of the first machine learning initiatives, A system that can play inspectors1955.