The Google Translate Journey: From Statistical to Neural to LLM | Macduff Hughes

Macduff Hughes spent 35 years in tech — including 8 years managing the Adobe Acrobat team and 12 years leading Google Translate. He co-authored the landmark 2016 GNMT research paper alongside Jeff Dean and 29 others, overseeing an error reduction of up to 85% on some language pairs overnight. He retired in 2024 having guided the team through statistical MT, neural MT, and into the LLM era.

 

In this episode:

- How statistical machine translation actually worked — and why it had a ceiling

- The 2016 neural transition: the decision, the disruption, and the human side

- What 500 million users and 200 languages taught him about language access and equity

- Gender bias in MT training data — and how the team tried to address it

- What the LLM era means for dedicated translation products

- Honest advice for mid-career translators navigating the shift

 

Watch on YouTube: https://youtu.be/dwYgYj_Cmvg

 

Connect with Macduff Hughes: https://www.linkedin.com/in/macduff-hughes-98665a3

 

Localization Fireside Chat on LinkedIn: https://www.linkedin.com/company/localization-fireside-chat

Robin Ayoub on LinkedIn: https://www.linkedin.com/in/robinayoub

Blog: https://www.robinayoub.blog

N49Networks: https://www.n49networks.com

Subscribe on YouTube: https://www.youtube.com/@L10nFiresidechat

Book a call with Robin: https://calendly.com/robin-ayoub/30min

Credits
Host: Robin Ayoub, Founder, Localization Fireside Chat

Connect and watch more
Website: www.L10NFiresideChat.com
Spotify: https://open.spotify.com/show/5OoURgc29R31XPGzOWL9iX
Be a guest: https://calendly.com/robin-ayoub/localization-fireside-chat-podcast-recording
LinkedIn: https://www.linkedin.com/in/robinayoub/

Support the show
Share the episode with a friend. Subscribe on YouTube and Spotify. Leave a short review.

Disclaimer
Views are for education and discussion. Apply judgment in your own context.

Tags
#L10NFiresideChat #localization #hiring #recruiting #leadership #ai

Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.