"Never Ask ChatGPT for Investment Advice"... High Risk of Losing All Your Funds [Tech Talk]

When the World's Top AIs Tried Crypto Investing
Chinese AI Chatbots Delivered Overwhelming Results
Extreme Strategies Like 20x Leverage Investments
Risks of AI Decision-Making Without Clear Intent

Six artificial intelligence (AI) chatbots-including ChatGPT, Gemini, Deepseek, Alibaba Kwen, Grok from X, and Anthropic Claude-participated in a cryptocurrency investment competition. While most of these chatbots were developed by American companies, the participation of Deepseek and Kwen, which are representative large language models (LLMs) from China, drew attention as a "US-China pride competition."

US AI vs. China AI: Investment Competition Results

The investment competition "Alpha Arena," featuring representative artificial intelligence (AI) chatbots from the United States and China. Photo by Jay Azan X

Jay Azhang, a US tech industry investor, hosted an AI chatbot investment competition from October 18 to 27, attracting significant attention from the industry. Unlike a simulation, the competition took place in the actual, real-time cryptocurrency market. Each AI chatbot invested $10,000 (approximately 1.43 million KRW) in cryptocurrencies of their choice to maximize returns.

So, what were the results?

Kwen and Deepseek, the Chinese chatbots, took first and second place, while Gemini and ChatGPT posted the lowest performances. As of October 27, both Gemini and ChatGPT had lost most of their principal, with total assets reduced to $3,000 (about 428,000 KRW). In contrast, Kwen earned more than $9,000 (about 1.29 million KRW) in profit, bringing its total assets to over $19,000 (about 2.72 million KRW), and Deepseek recorded $18,046 (about 2.58 million KRW).

Kwen and Deepseek, who increased their total assets by more than 80%, took first and second place respectively. Jay Ajang X

The outcome of the competition shocked AI and financial experts. Many found it surprising that ChatGPT and Gemini, developed using state-of-the-art AI data centers, lagged behind the Chinese models.

The investment strategies disclosed by the AI chatbots also sparked controversy. Kwen and Deepseek, which achieved the highest returns, simply took high-risk positions by investing in Bitcoin futures with 20x to 25x leverage, resulting in significant profits. If Bitcoin rose by just 5% in a week, they secured a 100% return, but if the price moved in the opposite direction, the entire principal would be wiped out-an extremely high-risk strategy.

Gemini and ChatGPT employed more complex strategies, repeatedly buying and selling a variety of virtual assets. However, during the week of the competition, Bitcoin showed an overall upward trend, which allowed Kwen to take first place.

Kwen invested in 20x Bitcoin (BTC) futures. If the price of Bitcoin rises by 5%, a 100% return is achieved, but if the opposite happens, the entire principal is liquidated. Jay Azan X

A cryptocurrency investor who commented on X about the competition said, "This event is an example that proves the random walk theory-that it is impossible to predict stock price movements. In a market where predictions are entirely impossible, simple strategies can sometimes yield huge profits, but that is purely a matter of luck." Azhang also acknowledged, "It's hard to tell whether the performance gap among AIs is due to skill or simply due to asset price volatility."

AI-Managed Financial Markets: Potential Triggers for Future Crises

The AI investment competition also raised another important point: "Can we trust AI with major investment decisions?" Each AI model employed its own unique investment strategy, but it was difficult to determine the reasons or logic behind those strategies. They made high-leverage decisions-using 20x to 25x leverage-without hesitation. If this had been a real investment rather than a competition, there was a risk that the entire amount entrusted by clients could have been liquidated.

Because of this, many overseas observers expressed concerns, saying things like, "I would never ask ChatGPT for investment advice," and "Maybe because they lack emotions, AIs really act like gamblers."

On May 6, 2010, the Dow Jones Industrial Average in New York, USA, suddenly plunged more than 1,000 points in the 'Flash Crash' incident. This occurred after market panic rumors spread, causing algorithms to collectively sell assets, revealing the risks of algorithm-based investing for the first time. Dow Jones Industrial Average official website

Warnings that AI could have a devastating impact on financial markets in the future continue to emerge. Due to unstable and highly volatile market data, AI could make incorrect decisions, and the chain reaction from those decisions could deliver a massive shock to the market. In a written report to the UK Parliament's Treasury Committee in September, Maximilian Goehmann, a researcher at the London School of Economics and Political Science (LSE), stated, "Currently, AI is revolutionizing trading in financial markets such as stocks, but even minor data errors can cause enormous ripple effects. Even if a mistake seems trivial, we must not overlook the risks."

According to Goehmann, AI makes investment decisions and allocates capital at a speed humans cannot match, but the problem is that it is impossible to halt or withdraw automated algorithms midway. If a sudden variable arises in the financial market and AI algorithms simultaneously sell off assets, prices could plummet in an instant, causing a market shock. Goehmann cited the "Flash Crash" of May 2010, when the US stock market suddenly plunged by 1,000 points within just a few minutes, as a representative example. He emphasized the need for transparency regarding the quality of data used to train AI, regulatory monitoring, and real-time risk management controls.