OpenAI says new models translate COBOL and Fortran more accurately
OpenAI said its latest AI models were much better at understanding and translating old code: think COBOL, the stuff supporting more than 40% of online banking systems, 80% of in-person credit card transactions, and 95% of ATM transactions, and Fortran.
These languages are tough to update because they have less code and documentation available for AI models to learn from, and "documentation may be incomplete, original developers may have left the company, and important business rules may only exist in code."
OpenAI models accurate, human verification required
In recent tests, OpenAI's GPT-6 Astra handled 20 programming languages with an 87.4% success rate, while GPT-6.1 Sol nearly aced COBOL-to-modern language conversions at 97.9%.
But even with these upgrades, OpenAI says human experts still need to double-check everything to make sure nothing breaks, especially since banking software can be very sensitive.