Toward 2035: Five technologies shaping the next decade

Amid intensified global tech competition and rapid technological advancements, artificial intelligence (AI)-driven research and cross-disciplinary innovation are key drivers of economic and social transformation. Recognizing these shifts in the technological landscape, FTChinese.com and The Innovation have launched “Toward 2035: Five technologies shaping the next decade,” with the final selections set to be unveiled at the 18th FT Chinese Annual Forum.

This initiative invited submissions from multidisciplinary research teams and R&D institutions. Through public polling and expert review, five technologies with broad social relevance were selected through public polling and expert review, ranging from new AI computing architectures to adaptive robotics, ovarian-health monitoring systems, aerospace energy hubs, and precision cloning methods. Together, they highlight the cutting-edge interdisciplinary work of China’s emerging generation of scientists.

The collaboration between FTChinese.com and The Innovation represents a rare intersection of a globally renowned financial and business media outlet with a leading academic journal. This unique partnership bridges the gap between business and academic. It aims to establish direct communication channels between emerging research teams and experienced business leaders, thereby creating pathways for commercializing scientific breakthroughs. Moving forward, facilitating the transition from research to application and strengthening the links between R&D and market validation will remain critical objectives for both academia and industry. Below are the profiles of the five selected technologies, listed in no particular order.

Reimagine deep learning and large models—From algorithms to precision

As training data nears depletion, model deployment triggers a severe energy crisis, and chip architecture encounters innovation bottlenecks, the field of AI is calling for a profound paradigm shift. Recent representative studies such as BackSlash, GCT, GGI, and ALS-LoRA is propelling a profound transformation of the AI paradigm: it employs differential geometry and manifold learning to diminish model reliance on data, leverages information theory to achieve joint optimization of training and architecture in response to the energy crisis, and designs a new generation of meta-computational units and parallel systems to break through chip bottlenecks. This culminates in an end-to-end co-design of models, algorithms, and hardware, pushing AI beyond simple computational stacking toward a new stage driven by information constraints (research team: LM2 Team, NYU Shanghai, Tsinghua University).