Automated trading algorithms fueled by artificial intelligence have quickly taken over the financial world. Promising safe, reliable returns, AI-driven investment platforms like Wealthfront and Betterment have attracted millions of investors, and millions more people expect AI to represent the future of stock trading.

But are we overestimating AI’s abilities and potential value?

Other investment strategies

First, we need to consider the fact that stocks, bonds, and index funds aren’t the only possible investments you should consider. The secret to a safe, high-growth portfolio is diversification, and that means learning about and investing in many other assets.

Algorithmic trading strategies are applied to more advanced financial transactions, like AI-based futures trading, but there’s still an inherent limit. This doesn’t mean the power of algorithmic trading is any less, but it does mean AI trading is currently restricted to a few, finitely defined areas.

The dangers of excessive AI

We also need to look at the potential dangers associated with trusting AI to make stock trades:

  • Reactivity and flash crashes. Most AI operations react to specific incidents with specific strategies. For example, when a major stock dips below a certain point, the AI is programmed to sell. The problem is, if too many programs are designed to operate this way, it can result in a destructive feedback loop. Selling drives the price down further, which triggers more selling, which can cause drops throughout the market in a cascading effect. And this isn’t speculation; it already happened in the flash crash of 2010. Fortunately, modern stock trading algorithms were then refined to prevent this kind of thing from happening again — but they’re not perfect, and these reactive strategies could become a real danger.
  • Exclusive dependence and volume. Some individuals may ultimately put their full trust in AI stock trading and nothing else. This strategy may seem safe and hands-off for the individuals doing it, but putting a huge percentage of your portfolio into one investment method or strategy is inherently risky. It’s even riskier if a large percentage of Americans are tied up the same way. If AI trading suffers a major loss, and those individuals lose a major piece of their portfolios, it could have widespread economic effects.

The limits of human-created technologies

Though it’s easy to think of an AI system designed to trade stocks as a super-intelligent robot, the reality is it’s still a program created by human beings. Human beings decided when these bots should buy or sell, and used their own reasoning and available historical data to make those decisions.

Because the humans consulted a wide range of investors and studied the historical fluctuations of the stock market overall, it’s likely that the AI will make effective decisions (and ones that will maximize potential returns). However, it’s important to realize that algorithms are still restricted by the limits of human knowledge. There’s an upper bound to how effective and safe they can be, and considering even top economists fail to predict economic trends, there’s a significant potential for error.

The complexities of the economy

Our economy is an incredibly complicated machine. It’s dependent on thousands of different variables, some of which we can’t even pinpoint or define, and they work together in spectacularly complex ways. To add even further complexity, not all of these variables are objectively definable. They can’t be reduced to a mere number or formula, and instead depend on more subjective feelings like consumer confidence or the perception of value.

It’s impossible to create a single formula that accounts for all these numbers, trends, data points, and feelings. Even if it were possible, it would be ridiculously difficult to program an algorithm to follow it reliably.

The future of AI stock trading

So are we overhyping or overestimating the value and potential future of AI stock trading? Perhaps. While there are dangers to excessively relying on AI stock trading, it currently has a great track record, and it was created using the same expertise and same data that human advisors use to make recommendations. Accordingly, it’s no more dangerous than trusting a human advisor to make your investment decisions.

That being said, there’s no ceiling to how far AI can develop in the future. More advanced versions of trading platforms could end up gradually creating and improving themselves, and might finally gain mastery over the complex variables that have eluded human economists for decades.

Larry Alton is a contributing writer at VentureBeat covering artificial intelligence.