AI in trading: potential and limits
Organise information, test hypotheses and recognise what a model does not know.
Research support
AI tools can help organise documents, summarise reports and formulate questions. They are most useful when the user knows the data source and reviews the output. Convincing language is not evidence of accuracy.
The limits of prediction
Models may fabricate references, use stale information or ignore costs. Results on historical data may reflect overfitting, information leakage or the selection of favourable periods. No backtest guarantees future performance.
Human oversight and data protection
Define what the tool may do and keep decisions under supervision. Do not share passwords, access keys or unnecessary personal information. On this site, AI is an educational topic; there is no AI-powered automated order execution.