Part 3/12:
Meanwhile, DeepSeek, a major player in this race, has made significant breakthroughs. Their R1 model has caught up with OpenAI's capabilities, proving that the so-called "moats"—distinctive advantages—are not insurmountable. Even more astonishing is their R10 model, which exhibits the ability to teach itself autonomously without human intervention. This is a landmark achievement—an AI that begins to evolve its reasoning and problem-solving skills independently, akin to the self-play strategy that made AlphaZero the world's best go player.