A few prompts are all it takes to ask ChatGPT for recommendations on a stock, a market, or even an aggressive strategy to outmaneuver the market. “GAI opens up new possibilities. All finance players are closely monitoring this rather than keeping their distance,” confirms Alain Bokobza, Global Head of Asset Allocation Strategy at Société Générale Corporate and Investment Banking (CIB).
For several years now, algorithms have enabled professional investors to enhance their analyses by “digesting” unstructured data for them: text (publications, economic surveys), images (satellite views), or audio (executives’ remarks on TV or radio).
“This information is translated into positive or negative signals. For example, analyzing satellite imagery allows us to measure deforestation or a company’s carbon emissions and the underlying risks,” explains Marie Brière, Scientific Director of the FaIR (Finance and Insurance Reloaded) program on the digitization of finance at the Louis Bachelier Institute.
Real-Time Adjustments Sentiment analysis is driving new behaviors. In 2023, American researchers showed that companies are adapting their language to a “machine” audience by avoiding words with negative connotations in their communications.
Algorithms could make markets more efficient overall, as the analytical power of new artificial intelligence (AI) engines helps identify and correct anomalies: unnoticed positive or negative information, temporary price discrepancies for the same stock listed in Paris versus New York… Adjustments are happening faster than ever.
“Previously, traders relied on monthly production or consumption indices. Now, robots analyze real-time data like supermarket parking lot occupancy rates or container ship traffic across oceans. Information is incorporated into prices as it emerges. However, this data can be skewed—for instance, by cloudy weather,” notes Marie Brière. Yet, amid this flood of short-term information, investors must “distinguish noise from meaningful insights,” cautions Alain Bokobza.
Prospective but Prudent Just as Google’s DeepMind mastered Go through machine learning and defeated the world champion in 2016, investment banks dream of developing neural networks that can learn stock market dynamics from scratch through iterations.
“This type of work accounts for 30% of internships at banks on both sides of the Atlantic. But it remains exploratory and forward-looking,” shares Gilles Pagès, Head of the El Karoui Master’s Program in Probability and Finance at Sorbonne University.
Moreover, financial operators must be able to explain to clients the rationale behind investment decisions. This rules out using AI systems whose reasoning is a “black box.” Finally, as long as AI engines are prone to hallucinations, extreme caution is necessary.
“Even if a machine priced derivatives accurately to the cent 99 times out of 100, it wouldn’t be enough. You can’t ever output an absurd price—otherwise, you risk losing millions in seconds,” adds Gilles Pagès.
AI won’t be replacing traders anytime soon.
This analysis adapts insights from Philippe Bernard’s Le Monde article, exploring AI’s transformative role in finance and investment strategies while balancing its potential with inherent limitations."
.webp)
0 Comments