Propose dataset/task for next meeting

[billhowe]

Everyone will propose a dataset/task they want to evaluate on local vs. small vs. large models. Just two items:

• description of task (what is being predicted / evaluated?)
• source of data (check to make sure it's accessible to us)

At the next meeting, we'll run a derby and see who wants to work on what.

The default research question in each case is "how does performance vary between local, small, and large models?" There might be other questions on a case by case basis