Mikio L. Braun

ML and AI expert

Back to archive

Thread

5 tweets

2
My favorite ideas from this interview: (1) data quality is more important than models. If you have an okay model, rather invest time to improve data quality than tweak the model.
8
3
(2) don’t spend years building up a data infrastructure first. Yes, data is important, but you also need to learn what kind of data you learn by doing small ML projects.
4
4
(3) don’t just look for the problems best suited for ML, but also for what the problems of the business are. This is something that is popping up frequently lately. Don’t solve just any customer problem, but an important one.
2
5
(4) there is a “small data regime” (less than 10000 data points) where you can for example look at individual data points and discuss what right label is.