Extreme climate events (ECEs) like heavy rainfall and heatwaves significantly impact society, and climate change is altering their frequency. Generalised Extreme Value (GEV) distributions help quantify these ECEs and guide human system design (e.g., return value of extreme wind gust sets construction codes at specific locations). We train a machine learning (ML) model using GEV distributions to determine the sample size required to estimate return values with specific uncertainties. For ECEs like heatwaves (with negative GEV shape parameters), fewer samples are needed to estimate the return value with specific uncertainty than rainfall extremes (positive shape parameters). For the heatwave example, a sample size of more than 20 times the annual recurrence interval is typically required to estimate the return value to ±10% uncertainty. A 1-in-20-year heatwave requires 400 samples, equating to 20 different 20-year simulations. Achieving such quantities will require extensive climate downscaling efforts, potentially aided by ML-based downscaling methods.