摩根大通（JPMorgan Chase & Co.）近期测试了人工智慧模型自主配置资金的能力。基于过去二十年历史数据的模拟测试显示，表现最佳的AI投资代理系统每年回报率比传统的60/40投资组合高出0.7个百分点，且波动性更低，同时也击败了摩根大通自身的规则化市场机制模型。

此项实验展示了华尔街采纳AI的下一阶段趋势，即AI系统可能从辅助员工的角色，晋升为直接决定如何在不同市场间分配资本。这些AI代理利用OpenAI和Anthropic等模型，将市场划分为金发女孩、通货再膨胀、滞胀和避险四种状态，并据此动态调整股票与债券等资产的配置比例。

尽管模拟结果令人鼓舞，但摩根大通分析师警告，这些结果仅基于历史模拟而非实际投资，不应将其视为AI能持续超越市场的保证。他们强调，代理型AI必须植根于深思熟虑的资产配置流程中，而非天真地认为AI模型本身就能提供专业领域知识。

JPMorgan Chase & Co. has been testing the ability of AI models to allocate capital autonomously. Backtests spanning the past two decades show that the best-performing AI investing agent outperformed a traditional 60/40 portfolio by 0.7 percentage point a year with lower volatility, while also beating JPMorgan's own rules-based market regime model.

The experiment offers a glimpse of Wall Street's next phase of AI adoption, where systems move beyond assisting workers to making decisions on capital allocation across markets. Powered by models from OpenAI and Anthropic, these agents classify the market into four regimes—Goldilocks, reflation, stagflation, and risk-off—and adjust asset allocations accordingly.

Although the early results are encouraging, JPMorgan strategists caution that they are based on historical simulations rather than live investing. They emphasize that agentic AI must be grounded in a well-thought-out asset allocation process, rather than naively assuming that the AI agent itself can serve as the primary source of domain knowledge.