Father of reinforcement learning predicted end of large language models era

Richard Sutton – this is one of the fathers of reinforcement learning and Turing Award laureate. So he stated that the era of large language models is coming to an end. Next, in his opinion, comes the era of experience. And here’s why he thinks so.

In his opinion, large language models are a dead end. Real intelligence should learn from experience, not from data. And modern neural networks – this is only imitation of intelligence. They have no experience, don’t perform actions and don’t receive feedback from reality. Therefore they’re not capable of real cognition.

According to Sutton, humanity is creating a new form of life based on design, not biological reproduction. And we’re becoming witnesses of the transition from a world where everything is copied to a world where everything is designed.

Living beings are replicators, and AI are designers, he explains. We can create systems that will create other systems, and all this through construction, not copying. According to him, this is a new stage of the Universe’s evolution.

It turns out that Sutton looks at the current boom of large language models as a temporary phenomenon. The real breakthrough, in his opinion, will happen when AI systems begin to learn through interaction with reality, receiving feedback from their actions. Not through consuming terabytes of text, but through real experience. As living beings do.

There you have the view of one of the founders of modern machine learning on the future of technology. It turns out that large language models are not the finale, but merely an intermediate stage.