World Models

Alibaba Cloud is investing heavily in “world models,” moving beyond text-based LLMs. These models aim to simulate reality using multimodal data. Alibaba led a 2 billion yuan investment in ShengShu, a startup developing a general world model for bridging digital and physical realms, exemplified by their Vidu AI video tool. This strategy aligns with Alibaba’s broader push into embodied intelligence and AI grounded in physical space, signaling a significant shift in AI development.

.Nvidia‑backed video‑generation startup Luma AI announced a European expansion, opening a London office that will house about 200 employees—40% of its staff—by early 2027. The move follows a $900 million Series C round led by Saudi fund Humain, pushing valuation above $4 billion. Luma develops multimodal “world models” for AI‑generated video, targeting marketing, media and entertainment. London’s talent pool and its role as a gateway to Europe underpin the strategy, as the video‑generation market is projected to exceed $15 billion by 2030.

Yann LeCun, a pioneer of AI, is leaving Meta to launch a startup focused on “world models” – AI systems that understand the physical world. Meta will partner with the new venture, signaling the strategic importance of LeCun’s research. This follows a restructuring of Meta’s AI division and a shift in strategy led by new AI chief, Alexandr Wang. LeCun’s move reflects a divergence in vision from the industry’s focus on large language models, as he champions alternative approaches toward achieving artificial general intelligence.