Machine translation no worse than human translation by 2040. — Axel Boldt
Created by Pablo on 2012-10-01; known on 2041-01-01
- Pablo estimated 50% on 2012-10-01
- Pablo said “Context: http://math-www.uni-paderborn.de/~axel/predictions.htmlMT will likely reach near-human levels by 2040. However, I’m not so sure it will completely match human translators.” on 2012-10-01
- Pablo estimated 25% and said “Upon further reflection, I’m lowering my estimate.” on 2012-10-01
- gwern estimated 70% and said “machine translation is already pretty good even with no AGI; somewhat likely AGI will be created by 2040 likely establishing superiority then and there; regular data sets only keep growing; 30+ years of tweaking can do a lot.” on 2012-10-01
- RandomThinker estimated 35% and said “The biggest risk is that “human” translation performance falls as people economize and third world students labor + machines.” on 2012-10-02
- NancyLebovitz said “No worse than average human translation or no worse than the best human translation?” on 2012-10-10
- Pablo said “Nancy, I’d interpret it as “No worse than a professional translation, such as those one finds in books published by mainstream publishers,”” on 2012-10-10
- chemotaxis101 estimated 55% on 2012-10-13
- smijer estimated 80% on 2012-10-19
- Qiaochu estimated 50% on 2012-12-06
- themusicgod1 estimated 96% on 2016-10-10
- pranomostro estimated 75% on 2018-12-08
- pranomostro estimated 85% on 2018-12-08
- Baeboo estimated 78% on 2018-12-12
- Medea estimated 70% on 2020-09-11
- JoshuaZ estimated 75% on 2020-09-11
- Baeboo estimated 79% on 2020-09-12