90 day update: I gave several AIs money to invest in the stock market

90-Day Update: Evaluating AI-Driven Investment Strategies in the Stock Market

In an earlier experiment, I shared a venture where multiple artificial intelligence (AI) agents were entrusted with real-time financial data and capital to participate in stock market investing. After approximately four months, I am excited to provide an update on how these AI models are performing, offering insights into the evolving intersection of AI and investment management.

The Setup

Four months ago, I initiated a test by allocating funds to several AI-powered trading agents. These models were designed to focus on swing trading and long-term investing rather than day trading, leveraging their access to live financial data to make informed decisions. The core hypothesis was straightforward: with comprehensive real-time data and a strategic approach, these AI systems could outperform traditional benchmarks, such as the S&P 500.

Performance Overview

As of today, the results are promising, though nuanced:

• Overall Market Comparison: Since the start of the experiment in November, the S&P 500 has declined approximately 7%, providing a challenging environment for all investors.

• Model Performances: Out of the evaluated AI agents, five have managed to outperform the S&P 500 from inception. Notably:

• Grok: Maintained positive momentum for most of the period but recently relinquished its gains, ending the period still ahead of the market.
• Claude and Gemini: These models have demonstrated the strongest average performance among the tested agents.

• GPT-based Models: Interestingly, all models based on GPT technology—despite their advanced language understanding—have underperformed relative to the market.

• Returns Summary: Despite some models surpassing the S&P 500, only two have achieved positive net returns after costs and market fluctuations.

Reflection and Future Directions

Given that this experiment spans just four months, it remains an early-stage investigation. The positive performances of certain models are encouraging, suggesting that AI can, under specific conditions, generate alpha in stock investing. However, to draw more definitive conclusions, extended testing and refinement are essential.

Longer-term observation will help determine whether these AI-driven strategies can sustain their edge and whether they can adapt to changing market dynamics effectively.

Conclusion

This ongoing experiment underscores the potential — and the current limitations — of AI in financial markets. While early results are promising for select models, comprehensive validation over longer periods is critical. Continued research could pave the way for AI strategies to become valuable tools for investors seeking systematic, disciplined investment approaches.

For those interested in following this project or exploring similar initiatives, more information can be found at Rallies.ai/arena.

Note: Investing involves risk, and AI strategies are continually evolving. Always exercise caution and conduct thorough research before deploying any automated trading system.

---

Author’s Note: This ongoing experiment exemplifies the innovative intersections of technology and finance. Stay tuned for future updates and analyses as the project progresses.