How reliable is artificial intelligence (AI) when it comes to investment or trading decisions?

I asked this question to

AI, to get a real sense of it, and this is what I’ve got, a short summary, of course:

AI is an extremely powerful tool for analyzing data.

But it should not be trusted as a fully autonomous decision-maker in financial markets.

The most effective approach today is what many professionals call “human-in-the-loop”: a combination of artificial intelligence and human judgment.

The main aspect we all should be aware of is that Artificial intelligence systems learn from historical data. They detect statistical patterns and correlations and use them to generate predictions about the future.

If something unprecedented happens in the world (for example, a new pandemic or a specific political conflict), AI may not have a model to follow. Popular language models (like ChatGPT or “Gemini”) can make mistakes in numbers or make up facts – without a blink of an eye! Which means we have a huge problem – hallucinations, or AI hallucinations, as they call it.

Practically everyone who uses AI tools on a daily basis knows what I am talking about. It may look cool if you need to make an image, but when it comes down to numbers and figures, or even more important stuff, such as data or your money, all jokes aside.

Sometimes mistakes are made in the most primitive ways, like ordinary Excel tables and data, where you simply need to add rows. It often happens that AI deletes some data or adds rows of data that are not there.

And here comes the first valuable insight you probably already know – you yourself need to KNOW the topic rather well to be able to manage the data you are presented. To be able to analyse it and make a sound decision on what to do next. To be able to quickly notice errors and so on.

So, when it comes to trading or investing, I think that the biggest problem with AI is the lack of context.

AI is great at seeing what is happening with the price, but it doesn't always understand why. Which means: detecting patterns is not the same as understanding why those patterns exist.

For example, the sudden departure of a founder from a company may look like a simple statistical shift to AI, while a human understands its devastating impact on the brand.

Another important limitation of AI becomes clear when we examine reflexivity theory, a concept popularized by well-known investor George Soros. Reflexivity describes the feedback loop between investor perception and market reality. In simplified form, the process works like this:

Investor perception → influences market prices → changing prices alter economic reality → which then changes investor perception.

This means that markets are also a part of a cycle along with human behavior and perception. So, markets are not merely reflections of economic fundamentals. They are also shaped by human beliefs and expectations.

AI, in turn, is based on historical data → pattern detection → prediction. But here are some big problems that arise, like Investor behavior changes, political regimes change, liquidity regimes change, and so on. Besides, market participants themselves change the system itself. Market participants are not passive observers! They are active participants who react to prices, react to narratives, and react to the behavior of other investors, and on top of that, they additionally react to the recommendations of AI models.

AI is good at analyzing data, but reflexive markets are often dominated by political decisions, Central Bank policies, regulation, liquidity, and institutional decisions. These factors transform the market structure itself. For example, a market bubble is a classic “feedback loop”:

The price starts to rise – investors become optimistic – more capital flows in – the price continues its winning streak – optimism only intensifies it – until... the bubble bursts. AI sees the trend and the data, but does not understand WHAT makes this “bubble feedback loop becomes unstable”.

Which simply put is that AI models can often detect the upward trend but to recognize when the feedback loop itself is becoming unstable is far more difficult. Simply because that requires interpretation of psychology, narrative momentum, liquidity conditions, and systemic risk.

Another problem concerning reflexivity:

The market is not an inert physical system. If an AI starts buying a stock because it is "cheap," that buying can itself create confidence in other algorithms, driving the price up and making the stock "expensive." AI often fails to calculate how its own actions change the future data points it is trying to predict.

One of the most famous examples of algorithms “going into a loop” and inflating the price to absurd levels occurred with biologist Peter Lawrence’s book “The Making of a Fly.” Around 2011, two Amazon sellers used automated pricing algorithms that followed each other. Both algorithms looked at each other as a “reference price.” Result – The price climbed to $23,698,655.93

But automated systems reacting to each other created a runaway feedback loop.

This example illustrates a deeper principle:

Algorithms interacting with other algorithms can amplify instability in unexpected ways.

Why is all this so important for a trader and investor? If markets were just statistics, then AI would be the perfect investor 😉 But financial markets are not just data systems — they are also strategic systems.

From the perspective of game theory, each investor’s outcome depends on the decisions of other investors. This means that market behavior emerges from strategic interaction between participants.

In short, AI is a great tool and can help you in many ways. It can work with huge amounts of data, but it should not be trusted to make decisions. It has many shortcomings. AI often experiences hallucinations and makes mistakes “on a flat surface”.

And abow all, tha AI DOES NOT BEAR ANY RESPONSIBILITY! 

Only YOUR knowledge can help you understand the markets. AI can help you to detect patterns, but only humans can interpret them. AI is not a solution: it's only an instrument for those who already KNOW the basics.

That’s why I suggest taking my program courses: THINK before you DO! That way, you’ll learn how to work with data, how to interpret that data, and above all, become more confident about your financial decisions. And you are welcome to get private tutoring from me once you take this course via Personal Helpline. See you on the other side!

Stay tuned!

Agris