This is not investment advice. The author has no position in any of the stocks mentioned. Wccftech.com has a disclosure and ethics policy.
Jan Szilagyi, CEO and co-founder of software firm Reflexivity, has shared interesting insights about the use of AI for trading stocks. In his appearance on CNBC's Squawk Box earlier today, Szilagyi shared that top hedge funds are using AI software as an add-on to their existing tools, and added that AI has already made several correct stock market predictions this year.
The executive, whose firm develops analytical engines overlayed with large language model (LLM) interfaces, believes that with sufficient computing power and training, AI can overcome human intelligence regarding trading stocks.
Szilagyi started out his interview by sharing that Reflexivity's end goal is to "build an autonomous investment analyst." In the process of doing so, the firm is currently providing technologies that enable investors to consolidate data from different sources under a single platform for streamlined decision-making.
By asking the 'engine' questions, users can "rely on the engine to go and find the data and then run the analysis" and reduce the time taken for such operations from two hours to two minutes, according to the CEO.
He outlined that Reflexivity's current AI trading systems are designed to "find parallels from the past to give you a sense of what the price pattern going forward could look like." These 'parallels' can "based on 12 or 15 past episodes, the selection of securities, that will correspond to what you're looking for" with regards to the current environment or a set of economic or other parameters that could entice certain stocks to deliver superior returns over others.
Szilagyi's clients are "primarily top hedge funds who are using this as kind of an intelligent overlay on top of all of the different data sources they have access to," he shared.
He also explained in detail how AI got several stock market trends right in 2024 and missed others. Szilagyi outlined that "If you look at the market as a whole, so for example it very correctly called the peak in July, it very correctly called the false rally that we had right after that." After the peak, the AI "again correctly called the bottom" as well.
However, the executive added that his software "turned bullish little bit too early" ahead of the Federal Reserve's latest meeting, following which the bank guided two interest rate cuts in 2025 as opposed to the earlier four.
Yet, the Szilyagi doesn't believe that the model was wrong here either since the its outputs are "always probabilistic assessments." According to him, these assessments are essentially the model "saying okay, you know 70% confidence this is what's gonna happen. There's still the 30% where it doesn't happen."
When asked when AI could completely replace humans in trading stocks, Szilyagi, rather controversially, stated that "it's hard to say that there's something so special about human intelligence that we should be better at trading." He believes that "If you are able to give the system so much compute power and the intelligence continues to improve, at some point it's not crazy to think the systems will be better at trading than humans are."
This leads him to conclude that when it comes to supplanting humans, "the horizon we're probably looking at is somewhere between five to ten years." Szilyagi added, "It doesn't mean that it's going to replace it entirely," since some sectors, such as private markets, will be harder for AI to target due to the relative dearth of data for training.