ByteDance founder Zhang Yiming is drawing a hard line on AI shortcuts: the company won't use model distillation to catch up quickly — even if it means falling behind Chinese rivals in the short term.
At an all-hands meeting of ByteDance's Seed AI team last month, Zhang made it clear that distillation is a shortcut the company isn't willing to take. He told the team to accept short-term sacrifices in exchange for long-term goals.
Distillation is a technique that lets a new model learn from the outputs of a stronger model, saving training time and compute. It's a tempting way to close the gap. Internally, ByteDance knows that skipping it will make it harder for Seed's language models to catch up with DeepSeek, Kimi, and Alibaba's Qwen in the near term.
But ByteDance's bigger worry is political risk. The company has spent years under US government scrutiny over TikTok, even as TikTok's US data security business has already brought in outside partners.