A recent study led by Washington State University suggests that large language models (LLMs) like ChatGPT are not yet capable of replacing finance professionals. While the AI systems excel at answering multiple-choice questions, they struggle with more complex, nuanced financial tasks. The research, which analyzed over 10,000 responses from ChatGPT, Llama, and BARD to real-world financial exam questions, found that ChatGPT performed best, particularly in areas where broad concepts are involved. However, it showed significant inaccuracies when dealing with specific financial scenarios, such as assessing clients’ insurance coverage or tax status.
The study’s authors, DJ Fairhurst from WSU and Daniel Greene from Clemson University, caution against overestimating AI’s capabilities. “It’s far too early to be worried about ChatGPT taking finance jobs completely,” said Fairhurst. While ChatGPT can synthesize information well for broad topics, it struggles with idiosyncratic issues. Despite performing well in reviewing market trends and securities transactions, the models’ limitations mean they cannot match human expertise in more complex financial decisions.
Related topics:
OpenAI Delays API Release for Sora Video Model Amid High Demand
OpenAI Launches Phone Number for ChatGPT Access via Calls and WhatsApp
ChatGPT 5 Expected to Launch in 2025, Promising Major Advancements in AI