Xiaomi’s Luo Fuli says Claude Fable 5 is an interim step in AI’s evolution
Luo Fuli, head of Xiaomi’s MiMo large model team, described Anthropic’s recently launched Claude Fable 5 as an interim-stage product, despite its advanced coding and agent capabilities. Speaking at the Beijing Academy of Artificial Intelligence Conference, Luo attributed Fable 5’s improvements to continued scaling across parameter count, computing power, and data, including synthetic data. She noted that while current models excel at execution and experimental design, they still lag behind top researchers in proposing novel hypotheses. Luo also highlighted that language models are currently advancing faster than world models, particularly in recreating conditions for intelligence emergence within agent environments.
Xiaomi’s assessment of Claude Fable 5 as an interim step underscores a critical perspective within Asia’s AI landscape: the rapid pace of development means even groundbreaking models are quickly viewed as stepping stones. This outlook drives aggressive investment in R&D across Chinese tech giants like Xiaomi, who are not content to merely adopt Western innovations but aim to lead in foundational AI research. Luo Fuli’s emphasis on scaling parameters, compute, and data, including synthetic data, reflects the prevailing strategy for pushing AI capabilities, a strategy heavily pursued by Asian firms with access to vast data sets and increasing compute infrastructure. The focus on agent-generated synthetic data is particularly relevant, suggesting a future where AI systems are not just trained on existing information but actively create new, high-quality data for their own improvement, potentially accelerating the development cycle within the region.
Furthermore, Luo’s distinction between current models’ execution prowess and their limitations in hypothesis generation highlights a key frontier for AI in Asia. Companies are not just looking for tools that can perform tasks but for systems that can innovate and drive scientific discovery. This ambition is fueling significant talent acquisition and academic partnerships across the region, as institutions and corporations strive to bridge this gap. The parallel advancement of language and world models, with language models currently leading, also indicates where immediate commercial and research efforts are concentrated, particularly in developing sophisticated agent systems for digital environments before fully tackling the complexities of the physical world. This strategic prioritization could see Asian companies dominate specific AI application domains in the near term.
