ByteDance plans to launch an AI assistant similar to Meta Muse in the coming weeks, aiming to consolidate its advantage in the AI video generation field and accelerate its catch-up with top U.S. AI labs, with the goal of narrowing the gap within roughly three years. Zhang Yiming unveiled the plan at an all-hands meeting of the AI team, expressing confidence that ByteDance has what it takes to compete with companies like Anthropic and OpenAI—though he acknowledged it will take until at least 2029 or 2030 to reach their level. Zhang has devoted considerable energy to shaping the company's AI strategy. He pointed out that the pretraining dataset scale of the current Seed series of large language models is far smaller than that of top U.S. competitors, and stressed that the company should not over-rely on model distillation technology but should focus on long-term development. ByteDance plans to invest up to $70 billion in a bid to regain an edge in the AI field. The company currently holds a clear advantage in AI video generation and is training a large model with a parameter scale of up to 10 trillion, while actively recruiting and retaining AI talent. The Seed team is also developing technologies such as robotics AI and virtual reality, and has set up an independent safety and alignment department.