Evolution strategies, a long-established optimization technique, has been found to perform similarly to reinforcement learning methods in various tests. Unlike reinforcement learning, evolution strategies do not require complex reward functions or environments with explicit feedback loops. This makes them potentially more scalable and easier to implement. --- Why it matters: This matters because it could provide researchers with an alternative approach to training AI agents that is less resource-intensive and more straightforward to implement. Source: https://openai.com/index/evolution-strategies