Why is AI So Bad at Financial Trading When Computers Are Just Better at Maths

One thing that everybody would have assumed that Machine Learning/AI would have been good at, given it is very good at forcasting, anomaly detection and maths in general, is financial trading. The problem is that several studies have shown that AI running on some very powerful machines is no better than any given human trader.

It made me wonder, why is AI so bad at trading compared to what it should be?

There are several reasons why AI may not always perform as well as humans in financial trading:

  1. Lack of Contextual Understanding: AI models, including sophisticated ones like GPT-3 or GPT-4, lack a true understanding of context. They generate responses based on patterns they’ve learned from training data, but they don’t actually comprehend the content in the way humans do. This can be a significant disadvantage in understanding complex financial markets.

  2. Unpredictable Market Conditions: Financial markets can be highly unpredictable, especially in the short term. Sudden events, news, or geopolitical shifts can significantly impact prices, and it’s challenging for AI models to adapt quickly to such changes.

  3. Limited Training Data: AI models rely on historical data to make predictions. If a model is trained on a limited dataset, it may not have seen a wide enough range of market conditions to make accurate predictions in all situations.

  4. Overfitting and Generalization Issues: AI models can sometimes become overly specialised in the data they were trained on. This is known as overfitting. They may not generalise well to new, unseen data, which is crucial in the constantly changing landscape of financial markets.

  5. Lack of Common Sense and Intuition: Humans have a deep understanding of the world and financial systems that is difficult to replicate in AI. This includes common sense, intuition, and an ability to interpret non-quantitative information (like news events, geopolitical factors, etc.) that may influence market movements.

  6. Regulatory Constraints: Financial markets are highly regulated, and AI models may not always comply with all regulatory requirements. Compliance with legal and ethical standards is crucial in trading.

  7. Risk Management and Decision-making under Uncertainty: Humans can exercise judgment and make decisions under uncertainty, considering factors that are not easily quantifiable. They can also incorporate risk management strategies based on their experience and intuition, which can be difficult for AI models to replicate.

  8. Lack of Emotional Intelligence: While emotions can sometimes lead to irrational decisions, they can also provide valuable insights and gut instincts. AI models don’t have emotions or instincts in the same way humans do.

  9. Market Manipulation and Game Theory: Humans are often engaged in strategic behavior and game theory when participating in financial markets. Understanding and anticipating these behaviors can be challenging for AI models.

It’s worth noting that while AI can have advantages in certain aspects of trading (e.g., processing large amounts of data quickly), it’s not a one-size-fits-all solution. Some trading strategies may benefit more from human intuition and judgment, while others may benefit from algorithmic approaches. In practice, a combination of human expertise and AI tools is often the most effective approach in financial trading.


comments powered by Disqus