> As an academic mathematician who spent their entire life collaborating openly on research problems and sharing my ideas with other people, it frustrates me [that] I am not even to give you a coherent description of some basic facts about this dataset, for example, its size. However there is a good reason for the secrecy. Language models train on large databases of knowledge, so you moment you make a database of maths questions public, the language models will train on it.

Well, yes and no. This is only true because we are talking about closed models from closed companies like so-called "OpenAI".

But if all models were truly open, then we could simply verify what they had been trained on, and make experiments with models that we could be sure had never seen the dataset.

Decades ago Microsoft (in the words of Ballmer and Gates) famously accused open source of being a "cancer" because of the cascading nature of the GPL.

But it's the opposite. In software, and in knowledge in general, the true disease is secrecy.