The emergence of large language models has prompted a surge of interest into theoretical models that might give us insight into both their successes and their shortcomings. We’ll give an overview of recent work in this direction, focusing on a surprising line of positive results that shows it is possible ...
