DeepMind：元强化学习揭示多巴胺在学习中的关键作用

Artificial Intelligence (AI) has dominated many video game leaderboards, such as Atari’s Breakout and Pong. Despite their success in video games, AI systems still rely on extensive game-specific training to match or surpass human performance. In contrast, humans quickly learn the basic gameplay of a novel game that they have never played before.

This discrepancy in learning speed and flexibility between humans and AI systems has led DeepMind researchers to investigate the role of dopamine in learning. Dopamine is a neurotransmitter in the brain that is known to play a role in reward, motivation, and attention.

In a recent study, DeepMind researchers developed a meta-reinforcement learning framework to explore the role of dopamine in learning. They found that dopamine plays a crucial role in learning by helping the AI system to focus on the most relevant information in the environment.

The researchers trained their AI system on a variety of tasks, including a simplified version of the Breakout game. They found that the AI system with dopamine-based learning outperformed the AI system without dopamine-based learning on all of the tasks.

The researchers also found that dopamine-based learning helped the AI system to learn more efficiently. The AI system with dopamine-based learning required less training data to reach the same level of performance as the AI system without dopamine-based learning.

These findings suggest that dopamine plays a key role in learning and that dopamine-based learning can be used to improve the performance of AI systems.