Hacker News new | past | comments | ask | show | jobs | submit login

I'm surprised they weren't already fully neural. NTM (neural machine translation) has been much better than everything else for a while now.



I'd imagine part of it is just having to test all the edge cases-- for every language is their full neural better?

That, and also building a good pipeline for it to work at scale.




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: