Patronus AI, founded by Anand Kannappan and Rebecca Qian, is addressing this issue by automating the testing of large language models (LLMs) with software. Despite the challenges, the co-founders remain optimistic about the potential of LLMs to assist in the finance industry, but stress that human involvement is currently necessary to ensure accuracy and reliability.
Key takeaways:
- Startup Patronus AI has identified limitations in AI models when analyzing SEC filings, with the best-performing AI model only achieving a 79% accuracy rate.
- One of the main issues is the AI models' tendency to refuse to answer questions or provide incorrect information not present in the SEC filings, posing significant concerns in regulated industries where precision is crucial.
- Patronus AI aims to address the challenge of the nondeterministic nature of large language models (LLMs) by automating LLM testing with software.
- Despite the identified limitations, the co-founders of Patronus AI remain optimistic about the potential of LLMs to assist professionals in the finance industry, with continued improvement.