An in-depth look at the 2017 “Attention Is All You Need” paper, a big breakthrough in AI, and profiles of the eight Google researchers who co-authored the paper
They met by chance, got hooked on an idea, and wrote the “Transformers” paper—the most consequential tech breakthrough in recent history. X: @jsmarr , @webbarr , @sandraupson , @hkanji , and @stevenlevy . Forums: Hacker News X: Joseph Smarr / @jsmarr : Great retrospective that lines up well with the people I knew and things I saw inside Google while it was developing. Those of us working on NLP paid a lot of attention to Jakob's research, but it wasn't until BERT that the whole company realized what a big deal this would be. @webbarr : Good read. Half is Greek to me but what's not is that of the 8 authors of Google's revolutionary transformer paper, none are still there... instead all but one are building their own (so far, v successful) companies & the one who didn't start a company is building a... Sandra Upson / @sandraupson : The biggest AI breakthrough in recent history was the invention of Transformers, at Google. @stevenlevy dug up the story behind that paper - how the researchers found each other, how they collaborated, and ultimately why they all quit. https://www.wired.com/... Hussein Kanji / @hkanji : They met by chance, got hooked on an idea, and wrote the “Transformers” paper—the most consequential tech breakthrough in recent history https://www.wired.com/... Steven Levy / @stevenlevy : The story of transformers. I talked to all eight of the authors of “Attention Is All You Need” the paper that revolutionized AI. https://www.wired.com/... Forums: Hacker News : 8 Google Employees Invented Modern AI. Here's the Inside Story
Context & Ripple Effects
This retrospective extends earlier coverage of how Google’s Transformer model accelerated machine language understanding by focusing on the researchers and institutional setting behind it. It also follows prior profiles of the same eight former Google scientists.
The personnel arc matters: none of the eight remains at Google, and all but one went on to start companies. That links a foundational research result to the dispersion of the people who developed it.
First-order effects
- Google no longer retains the team behind the paper; the authors’ departure has shifted their expertise and company-building activity outside the original lab.
- The account reinforces Transformers as the shared technical foundation connecting the authors’ Google research to later work in NLP, including the field disruption associated with LLMs and ChatGPT.
Second-order effects
- Research labs competing for frontier AI talent face a clearer retention challenge: a breakthrough can create new startup paths for the researchers who produce it.
- The spread of former lab researchers into new companies broadens the set of organizations able to commercialize and extend techniques that began inside a large platform company.
Third-order effects
- If this pattern persists, frontier AI advantage will depend less on a single lab’s ownership of a breakthrough and more on its ability to retain, recruit, and productize the teams behind it.
- The story is an example of frontier AI institutionalization: foundational research increasingly becomes an ecosystem-wide capability as talent moves between corporate labs and startups.
The trend: AI’s leading research breakthroughs are increasingly followed by talent diaspora, turning internal lab advances into a wider startup and industrial ecosystem.