[08] He He - Sequential Decisions and Predictions in NLP
The Thesis Review
The Future of Translation Models
This seems like a really interesting potential area for sequential decision making or reinforcement learning. It's definitely been making huge progress, but the main progress seems to be coming from the fact that the baseline MT system is so good right now. And even if you give it a partial sentence, it could translate that in a reasonable way. Although there is a risk that the model, we think the model is translated very well because it's fluent, but it's possible that it's just hallucinating some information to make the sentence look front, whereas the semantics are not correct.
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.
