I think we will continue to be surprised by how effective GPT-X and other LLM models are at coding tasks. The main reason is that unlike math programming is a f

I think we will continue to be surprised by how effective GPT-X and other LLM models are at coding tasks. The main reason is that unlike math programming is a field where dozens or hundreds of correct solutions exist for every task. All an LLM has to do is generate one out of hundreds of potential programs which solve the problem. It doesn't even have to generate perfect code, lots of profitable businesses are built off of buggy software. What's wild about this is that for now ChatGPT seems to have the most benefit for 'stack overflow' programmers who rely on copy-pasting solutions to problems other people wrote. Now they can effectively generate a stack overflow answer to their specific question instantly! For now AI seems like it will empower them.