While teaching an NLP class at Stanford, the speaker saw dozens of students each year use Twitter data to predict the stock market and present apparently successful results; the speaker says data leakage was always involved, and the task became less easy once leakage issues were fixed.

“Like I can tell you when I was teaching at Stanford, the NLP class,”

“like so many dozens every year said,”

“when I use DatasetX like Twitter to predict the stock market and they all like showed cute little things that somehow look like they were”

“and yeah and there's always some kind of data leakage and so on and it's just like wasn't as easy as they thought it would be once you once you fix all those issues but no I agree with you it's a very sensible application of AI yeah”