Top Jane Street Technologist Debunks ChatGPT Programming

Artificial intelligence (AI) has been touted as a threat to many jobs, including software engineering. However, Richard Eisenberg, a senior financial technologist and PhD open-source contributor at Jane Street, disagrees that AI-assisted programming is a sea change that removes the need for precise communication. In a recent podcast, he discussed the limitations inherent in programming languages that make chat GPT-powered programming challenging. Eisenberg suggested dependent types, an innovative technique he implemented for Haskell, can help ensure code generated by AI meets specified criteria. Used with AI, dependent types can create a prompt that tightens the reins and ensures the code meets specific aims. Eisenberg notes that dependent types are not a perfect panacea and can significantly impede speed, a setback in a company where speed is essential. He also warns against relying entirely on a computer’s coding abilities, insisting humans must maintain expert knowledge of function operations to ensure correctness.