Bard vs. GPT – Math, Logic and Problem Solving Battle

Let’s put the two AI’s to the test! … … with 10 multiple choice questions from the 2022 Australian Mathematics Competition.

I chose the Australian AMC questions as it is unlikely they are part of either model’s training data. (The AMC from the USA, on the other hand, may well have been used to train the models — OpenAI did state they were used in the evaluation of GPT-4.)

I'd like to see a yearly benchmark for models, could be logic puzzles or a suit of tasks but as it stands there is not good way to measure the ability of models.