At the start of last June, Berkshire Hathaway’s stock price suffered a spectacular 99% drop. The nosedive wasn’t caused by legendary investor Warren Buffet allocating his entire portfolio to NFTs, but by a technical glitch that also affected 40 other stocks, including miner Barrick Gold, which briefly traded for 25 cents per share, compared to the$17 before the glitch.

Of course, technical failures and errors happen often, and they always will. Most of them go unnoticed by the majority of the population, while others cause massive disruption, like the recent Microsoft/CrowdStrike fiasco, which grounded flights all over the world and paralyzed countless businesses. Clearly, global systems integration and centralization greatly amplify the risk of large-scale impact; if everyone is using the same basis to build complicated systems on, a domino effect is unavoidable. This is precisely why these risks are particularly worrying in the financial sector and in the investment industry.

The rise of algo-trading, meaning the use of specialized computer programs to make investment decisions with the kind of speed and efficiency that no human could ever rival, has already been problematic in this regard and it has substantially exacerbated market fluctuations. This is because even though these proprietary algorithms have different parameters and goals built into them, they all draw raw data from the same sources. It’s easy to see how one faulty input can make many of them go haywire and how a tiny mistake can cause a butterfly effect.

This risk is now being elevated to a whole new level with the introduction of AI technology into the mix. For one thing, the technology is very far from perfect at this point. As we’ve seen from their chatbot peers, like ChatGPT or Google’s Gemini, even their latest versions still have a tendency to “hallucinate”, i.e. to come up with totally made-up information and data to answer users’ questions.

Instead of “admitting” they don’t know the answer, they just invent facts and present them with staggering confidence, leaving the user unable to differentiate between fact and fiction. Earlier this year, ChatGPT reportedly landed lawyers in hot water in court as they used the AI platform to prepare their arguments and the chatbot fabricated case law. They fail at even more basic tasks, like elementary natural history: when asked about prehistoric civilization, ChatGPT asserted there are fossil remains of dinosaur tools and that "some species of dinosaurs even developed primitive forms of art, such as engravings on stones".

It is obvious that it would be incredibly risky for any investor to rely on these tools at the moment. However, progress is moving along at a lightning fast pace in the AI field, which is being flooded by fresh investments and has attracted the most brilliant developers from all over the world. Therefore, it is safe to assume that one day, in the not-so distant future, we will have AI investing tools and programs that are absolutely fit for duty. That doesn’t mean that the problems will end there though.

The trouble with the widespread adoption of AI in finance is that it could lead to (even more) market volatility and amplify crises much more than the simple algorithms of old do. AI systems draw data from more sources and in a very different way. They “learn” as they go and they evolve, perhaps too quickly and too unexpectedly for humans to keep up with. For instance, a sudden shift in market sentiment or an unexpected economic event could trigger cascading selloffs by AI trading systems, much like we see today with algo-trading, but on a much bigger scale. That NYSE glitch could have been a whole lot worse if AI investing was more widespread. Another concern over introducing AIto investing is the potential bias. AI models are only as good as the data they're trained on, and if this input data is flawed or skewed in any way, the output be massively problematic.

That’s not to say that AI has nothing positive to bring to the table. On the contrary, it can massively increase efficiency and improve risk management by analyzing complex data and scenarios. However, all new technology comes with caveats, and it is important for investors to fully understand the risks. Even if you’re not personally using these tools for your investment decisions, hordes of other investors- institutional ones too - will be using them, so you must expect a great shift in the market environment and conditions in the coming months and years.

This article is part of our BFI Bullion Digger Newsletter! Find the full Digger here. The other two articles discuss the questions our BFI Bullion team has been asked the most this year, and takes a look at how banks are truly doing. It explains how many people have taken bank stability for granted for way too long.