The Google Translate Journey: From Statistical to Neural to LLM | Macduff Hughes
The Google Translate Journey
Listen to the episode
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…
Watch the full episode
Shorts from this episode

Gender Bias in Machine Translation — And How Google Addressed It | LFC Ep. 223

The Future of Conversational Translation — And Why It's Harder Than It Looks | LFC Ep. 223

80% of the Web Is in Just 10 Languages — What That Really Means | LFC Ep. 223

How LLMs Changed the Way Macduff Thinks About Translation Products | LFC Ep. 223

What Mid-Career Translators Need to Hear Right Now | LFC Ep. 223

The Second Paradigm Shift: Leading Through the LLM Era | Macduff Hughes | LFC Ep. 223

The Moment the Team Knew Neural Networks Were the Answer | LFC Ep. 223

Is the Translation Industry Making the Right Bets? | Macduff Hughes | LFC Ep. 223

Was Google Translate a Revenue Product? The Real Answer | LFC Ep. 223

The Accidental Path to Google Translate | Macduff Hughes | LFC Ep. 223

How Statistical Machine Translation Actually Worked | Macduff Hughes | LFC Ep. 223

Why Governments Were Lobbying Google Translate | LFC Ep. 223

The Data Problem: How Google Translate Added New Languages | LFC Ep. 223

35 Years in Tech: What Macduff Hughes Would Tell His Younger Self | LFC Ep. 223

From 40 Languages to 200 — What It Took | Macduff Hughes | LFC Ep. 223

What the Tech Industry Gets Wrong About Language | Macduff Hughes | LFC Ep. 223

Why Macduff Hughes Started Learning Welsh | LFC Ep. 223

Error Rates Dropped 85% Overnight — Here's How | LFC Ep. 223

The Parallel Between PDF Infrastructure and Google Translate | LFC Ep. 223

What the 2016 Google Translate Switch Actually Looked Like | Macduff Hughes | LFC Ep. 223
