Chinese Quant Fund High-Flyer Capital Challenges AI Giants with New Model
According to the Financial Times, High-Flyer Capital Management, a prominent Chinese quantitative hedge fund, is making waves by leveraging its AI expertise to challenge AI giants such as OpenAI and others with its new model, DeepSeek-V2.
High-Flyer’s new AI model, DeepSeek-V2, launched in May, is already garnering significant attention due to its ability to answer questions, write code, and reason. This positions it as a strong competitor to leading Western AI models such as GPT, Llama, and others.
But what sets DeepSeek-V2 apart is its cost efficiency, charging just Rmb2 for every million output tokens, a pricing strategy that has sparked a price war among Chinese AI providers. Shortly after DeepSeek’s release, technology giant ByteDance slashed its prices to Rmb0.60 per million output tokens.
In response, Alibaba reduced usage prices for some models by up to 97%, while Baidu made two of its Ernie models free. This intense competition shows that China’s AI market has evolved and become quite dynamic and highly competitive as the benefits of generative AI are found.
Liu Qingfeng, founder of Chinese AI group iFlytek, emphasized the narrowing gap between the US and China in AI. “The gap between the US and China isn’t as big as everyone thinks,” Liu stated at a recent tech gathering in Macau. “In a lot of verticals, our [models] are better than theirs.”
High-Flyer’s journey began in a modest Chengdu apartment, where founder Liang Wenfeng, a computer science graduate from Zhejiang University, experimented with automated stock trading. Today, the fund’s strategies are heavily reliant on AI, akin to the approaches pioneered by Renaissance Technologies. “AI helps to extract valuable data from massive data sets which can be useful for predicting stock prices and making investment decisions,” explained manager Cai Liyu during a 2021 online roadshow.
DeepSeek’s development is significantly bolstered by High-Flyer’s financial success. The hedge fund has returned 151% since 2017, a remarkable achievement given China’s volatile stock market which has been shaken by real estate and other issues.
But despite recent regulatory challenges, including a crackdown on quant funds by Beijing, High-Flyer has been able to remain resilient. High-Flyer’s substantial investments in AI infrastructure have been crucial to DeepSeek’s success. The company has built a second supercomputing cluster, connecting over 10,000 Nvidia processors, enabling the training of large AI models.
This investment was made before US restrictions on advanced chip exports to China took effect in mid-2022. DeepSeek-V2’s performance has been impressive. Research groups, including the University of Waterloo, rank it among the top 10 language models globally, trailing only behind OpenAI’s GPT-4, Anthropic’s Claude, and Chinese rival 01.AI.
“The model’s architecture is…unique,” noted Andrew Carr, chief scientist at US-based AI animation start-up Cartwheel. “DeepSeek has taken this idea called mixture of experts, where you split up a model into smaller chunks, to the extreme, with hundreds of small experts.”
Carr added that DeepSeek’s pricing is about 100 times cheaper than OpenAI’s GPT-4 and a fifth of Anthropic’s Claude 3 Haiku. Inside China, DeepSeek’s competitive pricing has quickly attracted developers. “I switched from using OpenAI’s GPT-4 for coding help to DeepSeek because of the lower prices,” said Wang Zixu, a programmer based in northern China.
Despite the cost advantage, some experts suggest that DeepSeek might be operating at a loss at its current price point. Additionally, the model’s computing power could lag behind as newer, export-restricted Nvidia chips become available.
High-Flyer and its AI offshoot are ambitiously aiming for artificial general intelligence (AGI), where machines surpass human cognitive abilities. A recent DeepSeek job advertisement captured this vision: “We believe AGI is the violent beauty of model x data x computing power. Embark on a ‘deep quest’ with us on the journey towards AGI!”
Originally posted on OpenDataScience.com
Read more data science articles on OpenDataScience.com, including tutorials and guides from beginner to advanced levels! Subscribe to our weekly newsletter here and receive the latest news every Thursday. You can also get data science training on-demand wherever you are with our Ai+ Training platform. Interested in attending an ODSC event? Learn more about our upcoming events here.
