Borrowing data from OpenML (or HuggingFace and Kaggle)

[BradKML]

Would it be possible for the AI Scientists to get data from established sites so that research can be accelerated?
Problem 1: they might require use logins to use their APIs (e.g. Kaggle or some of HF)
Problem 2: the data has to match the task
Problem 3: for some datasets, the "X and y" can only be known in the documentation

• https://www.openml.org through https://www.openml.org/apis
• https://huggingface.co/datasets through https://huggingface.co/docs/hub/en/api
• https://www.kaggle.com/datasets https://www.kaggle.com/docs/api

Bonus observation: https://arxiv.org/html/2403.19546v3