This is probably the best thing OpenAI could have done to make sure ChatGPT had access to a treasure trove of curated scientific data.
Might be overkill for fixing its issues with simple addition and multiplication, but will definitely limit factual hallucination and make many kinds of answers more trustworthy.