JPMorgan Patent Aims to Verify AI-Generated Code
JPMorgan is seeking to patent a method to double-check the quality of software code generated by Large Language Models (LLMs), ensuring accuracy and reliability.
AI can always write for you, but it’s not always right.
The proposed system works by having a user provide instructions and desired output, which are then sent to the LLM along with available APIs. The LLM selects an API and generates code, which is then run and evaluated for quality, including accuracy, reliability, and consistency.
- User provides instructions and desired output
- System sends information to LLM along with available APIs
- LLM selects an API and generates code
- System runs and evaluates the code for quality
This is not the first time JPMorgan has sought patents to ensure accuracy, having previously explored the use of machine learning and blockchain for fact-checking.
Author's summary: JPMorgan seeks patent for AI code verification method.