logo

Chinese retail traders embrace DeepSeek in a nod to quants

Zawya11-03-2025
SHANGHAI/SINGAPORE - If you cannot fight them, join them, is the mantra among Chinese mom-and-pop investors who are embracing DeepSeek and other artificial intelligence tools, in sharp contrast with last year's government crackdown on computer-driven quantitative traders.
Online crash courses have mushroomed and training rooms are packed with retail traders eager to beat the market using computer models, as the popularity of DeepSeek - itself backed by a quant fund - changed not only the market trajectory, but the perception of China's $700 billion hedge fund industry.
The rapid adoption of DeepSeek in China's retail-dominated stock market is also prompting changes at brokerages and wealth managers, while creating new risks for investors in a market dominated and driven by small-time traders' cash flow.
"The future is the digital age, and AI will be vital," Hong Yangjun told a packed room of individual investors learning to trade with AI on a weekend in February.
Just as future warfare will be fought with drones and robots, the stock market will be a battleground between computers, the lecturers told the class in an office in downtown Shanghai.
Such piety is in stark contrast to the public outcry a year ago against computer-driven quant funds, viewed as "bloodsuckers" by retail investors, and blamed by regulators for contributing to market unfairness and volatility.
The industry was also the target of a government crackdown roughly a year ago, when the sector was worth $260 billion by some estimates.
Last month, however, investors handed over 15,800 yuan ($2,179.91) each for a weekend lecture by Mao Yuchun, founder of Alpha Squared Capital, on how to trade stocks with AI, according to the organiser, who promoted the event by drawing attention to Alpha Squared's geographical affinity with High-Flyer.
High-Flyer, based in eastern Hangzhou, is the hedge fund behind DeepSeek - the Chinese AI start up that stunned Silicon Valley with its cost-efficient large language model and spurred a rally in Chinese stocks.
Meanwhile, Chinese social media is brimming with online courses teaching traders how to use DeepSeek to evaluate companies, pick stocks, and code trading strategies.
"Using quantitative tools to pick stocks saves a lot of time," said Wen Hao, a Hangzhou-based trader.
"You can also use DeepSeek to write codes," said Wen, who uses computer programs to determine the timing for buying and selling.
U.S. fund giants including BlackRock, Renaissance Technologies and Two Sigma have already been using AI in investing for some time. Analysts say small asset managers and even retail investors in China stand to benefit from the emergence of DeepSeek's open-sourced model.
ChatGPT is off-limits to Chinese users.
DEEPSEEK ADVICE
The fondness for the AI-led turnaround in the perception of quant trading has coincided with a sunny start to the year for stocks, after a few years in the doldrums.
Goldman Sachs said the MSCI China index has made its best start of the year in history and brokers are racing to build AI models into their platforms.
"In the future, Chinese investors will completely change the way they make investment decisions and place orders," said Zhou Lefeng, president of Xiangcai Securities.
"Previously, clients would ask wealth managers for investment advice. Now they ask DeepSeek."
Larry Cao, principal analyst at FinAI Research, said DeepSeek is popular because it's cost-efficient, has strong reasoning ability, and unlike ChatGPT, is readily available, and is promoted by the Chinese government.
Nevertheless, he is surprised at the level of faith investors put in the model, cautioning that AI has limits.
"People trust AI models more than they trust financial advisers, which is probably misplaced trust at least at this stage," Cao said.
There could also be a herding effect if one school trains many retail investors to trade using the same signal.
"Large language models seem impressive. But at this stage, they are not necessarily smarter than most investors."
What's certain, said Feng Ji, CEO of Baiont Quant, is that DeepSeek has changed retail perception of quant fund managers.
"I can feel strongly that the public are thinking twice about quant fund managers' contributions to society," said Feng, whose company uses machine learning to trade.
"I never think we caused retail investors' losses. We actually provide liquidity and make the market more efficient." ($1 = 7.2480 Chinese yuan renminbi) (Reporting by Samuel Shen and Tom Westbrook Editing by Shri Navaratnam)
Reuters
Orange background

Try Our AI Features

Explore what Daily8 AI can do for you:

Comments

No comments yet...

Related Articles

Why AI can't replace human creativity and never should
Why AI can't replace human creativity and never should

Khaleej Times

time7 hours ago

  • Khaleej Times

Why AI can't replace human creativity and never should

I have been negative about AI, or rather the large language models and hive-minds of underpaid desk-jockeys and migrants on computers forced to masquerade as chat-bots (look it up), while also being certain it will be worse for humanity than social media. Those fears are bearing out with people using LLMs as therapists who don't have medical degrees and study pals who will lie and tell you what you want to hear if it keeps you engaged. Again, look it up, and do not use the same search engine unless it is to find another. This isn't widely known or even understood, again thanks to social media chopping up society into little bite-sized pieces to get gobbled up by the endless churn of rich and powerful surrounded by sycophants seeking more money and power to fill a hole inside themselves. But it is a fact that these so-called artificial intelligences are being trained on stolen material, and generate their work based on that theft. In the UK, this has reached a head with the government doubling down on plans to make it easier for these companies to steal original material, where the creator does not know to what extent they need to 'opt out' of having their work stolen. This in turn is due to a lack of AI transparency laws, but the fact of the matter is if I tell an LLM to make me a picture of one thing in the style of a specific artist, I know that the AI is going into its database of that artist's work to generate a composite of it and others into what I want made. Because it isn't drawn, it is generated. Drawings, or writing for that matter, has stages. You pause, you edit, you back up and readjust based on feeling and emotion. You do all the things humans do about everything, primarily over-think, just to be dissatisfied at the result. I experienced this first hand and even asked one LLM to try writing a column for me; it was not bad, but it was far from good because it had no life. I'm Gen-Z and a proud digital native, and there's a certain way of communicating that only people who grew up online have; that is, we talk in the exact same tone as we write. For me, just without the stutter. In needing to emerge and find friends online, that connection between what I would say and how I would write it, is short. I can self-edit very well, and what comes to matter most is ensuring I am understood, rather than plastering words together emotionally, and if I feel good about it too bad for everyone else. That is what all LLMs do; take the lowest common denominator of what was asked for and spit it out, because that is what its purpose is. Truly, LLMs are only digital machines; they would only be intelligent if they thought for themselves, which they do not, because they are incapable of recalling on past experiences or accessing memory through emotional triggers. In the case of one famous chat bot everyone is using far too much, it will save a response a user liked to be regurgitated at a later time for another user, with some teachers knowing their students used AI on an assignment by the similarity of individual sentences and passages across different students' assignments. The only thing left to do is figure out a way to secure creative product from being sucked into a training model, while also leaving it free for dissemination among the people it is intended for.

Flawless closes $1.5mln pre-seed funding to expand business
Flawless closes $1.5mln pre-seed funding to expand business

Zawya

time9 hours ago

  • Zawya

Flawless closes $1.5mln pre-seed funding to expand business

Flawless, a Saudi digital platform offering AI-powered career guidance, has secured $1.50 million in a pre-seed funding round backed by a group of regional angel investors. The investment will be used to advance the platform's technology, enhance user experience, and expand the company's growing team, according to a press release. Founded by Shaimaa Alghamdi, Flawless integrates generative AI and social psychology insights to offer tailored career recommendations to individuals navigating professional transitions. The platform originated as a blog in 2023 and transformed into a full-fledged digital startup in 2024, targeting unmet demand in the career support space. In line with the economic development of Saudi Vision 2030, Flawless quickly scaled by identifying a gap in the market for intelligent, tech-driven career guidance. The company aims to evolve its services and widen its reach across Saudi Arabia and beyond. Alghamdi commented: 'What began as a passion project is now a data-driven platform helping thousands make smarter career decisions. This funding validates our approach and gives us the fuel to scale responsibly and impactfully.'

Nvidia set to become the world's most valuable company in history
Nvidia set to become the world's most valuable company in history

Zawya

time9 hours ago

  • Zawya

Nvidia set to become the world's most valuable company in history

Nvidia was on track to become the most valuable company in history on Thursday, with the chipmaker's market capitalization reaching $3.915 trillion as Wall Street doubled down on optimism about AI. Shares of the leading designer of high-end AI chips were up 2% at $160.4 in morning trading, giving the company a higher market capitalization than Apple's record closing value of $3.915 trillion on December 26, 2024. (Reporting by Noel Randewich in Oakland, California; Arsheeya Bajwa in Bengaluru and Carolina Mandl in New York; Editing by Dawn Kopecki and Matthew Lewis)

DOWNLOAD THE APP

Get Started Now: Download the App

Ready to dive into a world of global content with local flavor? Download Daily8 app today from your preferred app store and start exploring.
app-storeplay-store