Artificial intelligence · 2016
Google Neural Machine Translation
Google Translate switches to neural networks and cuts errors by more than half.
GNMT used deep LSTM encoder-decoders with attention and was deployed for Chinese-English in September 2016. Side-by-side human ratings showed it cut translation errors by 55-85% on major language pairs, replacing a decade of statistical MT almost overnight.
What it built on 2
What it led to 2
Related measurements
The dashed line marks when it first worked.
Training compute of notable AI models Source: Epoch AI, Notable AI Models (CC BY 4.0)
ML hardware FLOP/s per dollar Source: Epoch AI, ML hardware (CC BY 4.0)