ChatGPT for Programming Numerical Methods

ChatGPT is a large language model recently released by the OpenAI company. In this technical report, we explore for first time capability of programming numerical algorithms. Specifically, examine GhatGPT generating codes algorithms in different languages, debugging and improving written users, completing missed parts codes, rewriting available other parallelizing serial codes. Additionally, assess if can recognize given are humans or machines. To reach goal, consider variety mathematical problems such as Poisson equation, diffusion incompressible Navier-Stokes equations, compressible inviscid flow, eigenvalue problems, solving linear systems storing sparse matrices, etc. Furthermore, exemplify scientific machine learning physics-informed neural networks convolutional with applications to computational physics. Through these examples, investigate successes, failures, challenges ChatGPT. Examples failures producing singular operations on arrays incompatible sizes, interruption relatively long Our outcomes suggest that successfully program but certain limitations exist require further improvement model.