I asked both Bard and GPT-4 to write a code that simulate a M/M/G queue with some simple additional information about request distributions. Bard generated an i

I asked both Bard and GPT-4 to write a code that simulate a M/M/G queue with some simple additional information about request distributions. Bard generated an incorrect implementation with an infinitely loop and without using any discrete simulation package. GPT-4 generated a less incorrect implementation that uses simpy. When I asked both models to convert normal distribution with a binomial distribution, Bard got confused and generated incorrect implementation of binomial distribution, while GPT-4 correctly replaced the normal distribution in the generated code with correct binomial generator.