The Gradient: Perspectives on AI cover image

Kyunghyun Cho: Neural Machine Translation, Language, and Doing Good Science

The Gradient: Perspectives on AI

CHAPTER

Scalability and Reason and Data Analysis - I'm Very Much Guilty

As the algorithms become increasingly more complicated and then also the number of hyperparameters associated with these algorithms grows, it becomes increasingly easier for people to exploit the fact that we only have a single benchmark or a fixed small number of benchmarks. Because how much effort you put into each algorithm can impact the outcome very much, because when you have a very small number of data points for testing them. We are in a interesting regimen where rigorous statistical analysis is very difficult, just because of the scalar which we work. And at the same time, we feel like we may not necessarily need it because we are looking at so many different aspects using so many data points at thesame time.

00:00
Transcript
Play full episode

Remember Everything You Learn from Podcasts

Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
App store bannerPlay store banner