Goku Technologies, a Shanghai-based quantitative trading fund, is expanding its ambitions beyond finance, looking to artificial intelligence (AI) as a means to overcome challenges faced by humanity, according to CEO Wang Xiao. In an interview with the Post, Wang said the firm’s AI initiatives were driven by a vision that transcended short-term commercial interests. “What I believe is [AI exists] to solve issues for humanity – this is the only thing that matters,” he said. Last month, Goku launched Shanghai AllMind Artificial Intelligence Technology, an AI-focused subsidiary, in a move that draws parallels with Chinese hedge fund High-Flyer Quant, which established DeepSeek in 2023. Wang, however, played down the comparison, pointing out that Goku was pursuing a different path with AllMind. Instead of competing directly with major tech companies like Alibaba Group Holding or start-ups like DeepSeek, which develop AI models from scratch, AllMind was envisioned as a catalyst for AI-assisted scientific discovery, according to Wang. Alibaba owns the South China Morning Post. Goku gained attention last month with the release of a research paper introducing a new AI training technique that combines supervised fine-tuning and reinforcement learning – key methods used by industry leaders like OpenAI and DeepSeek. In the paper, Goku and AllMind researchers claimed that their approach surpassed mainstream techniques used by leading AI research labs. The company, in collaboration with Shanghai Jiao Tong University, submitted its research to the Conference on Neural Information Processing Systems, often referred to as the “AI Olympics”. It marked the first time a Chinese quant fund had contributed to the annual event, which brings together top machine-learning scientists. Wang expected AllMind to incur annual costs of around 50 million yuan (US$7 million). While the unit is currently fully funded by Goku – one of China’s largest quant funds, with over 15 billion yuan in assets under management last year – Wang said the firm might seek external investment in the future. Goku’s significant push into AI is not a spontaneous side project. The fund established an internal AI team in 2017 – just two years after its founding in 2015 – to explore the use of algorithms and related technology in trading. While companies like OpenAI, Alibaba and DeepSeek have grabbed attention with consumer-facing generative AI products like chatbots, Goku has adopted a more research-driven approach. Researchers at Goku and AllMind focus on fundamental AI model research to explore how the technology can be integrated into quantitative trading as well as other fields, according to Wang. “I told my team that to use [AI models] well, we need to understand them inside and out, so we began working on topic-based fundamental research,” he said, referring to the company’s joint work with Shanghai Jiao Tong University on model training. Through that collaboration, the team discovered that its AI training infrastructure for processing financial data could also support research and development in areas such as new materials, chemistry and biology. “After we completed that research, our next step would be to use AI for science,” Wang said. Wang, who holds a doctorate in physics from Shanghai’s Fudan University, serves as CEO for both Goku and AllMind. He said the lines between the two roles had blurred, as his daily activities revolved largely around discussions with engineers about AI models. “Anything related to large language models is always at the top of my agenda,” he said.