The big picture: Research units within large tech companies have made many fundamental technical breakthroughs in the past. But as technology has played an ever bigger role in society, and more company-backed research has focused on tech's social impact, researchers' findings don't always line up with corporate objectives.
Details:
- Gebru, a leader in the field of AI ethics, wrote Google management with a list of concerns after the company insisted its name be taken off a paper she and others had submitted for an industry conference.
- The paper explores problems with large-scale language processing — Google's foundational business.
- Gebru said she'd leave Google if the company didn't address her concerns, and Google told her it took that as an immediate resignation.
- Separately, Gebru had also written an email that circulated widely within Google criticizing the company's record on diversity.
- Several thousand people, from both inside Google and elsewhere, have signed an open letter calling on Google to explain its actions and commit to greater transparency.
Between the lines: Gebru has been a strong advocate for reducing bias in AI, and a paper she co-authored exposed how gender and racial basis is built into many facial-recognition systems.
- Sources inside and outside of Google say her work in both ethics and diversity had ruffled feathers in some corners of the company even as she maintained the support of both her manager and the broader team focused on AI ethics.
- Meredith Whittaker, who helped organize last year's Google employee walkout, said that Google's move sends a chilling message. "When they felt she was pushing too hard on the issues she was hired to advocate for, they fired her," Whittaker said.
State of play: Within Google's research unit, the move has raised concerns about the company's commitment to academic freedom and suggested to some that women and people of color are targeted for retribution when they speak out.
- It also comes as Google finds itself under scrutiny from government, critics and its own workers for everything from allegations of monopolistic practices to its treatment of contractors to the business it does with military and government agencies.
- Known for many years for its freewheeling and outspoken culture, Google has more recently dialed back its toleration of internal debate and dissent.
The other side: in a public note that also included his Thursday email to his staff, Google AI head Jeff Dean noted that Google's research and publication practices are designed to reflect that it employs a wide range of researchers, some of whom disagree with one another.
- He says the company has "a strong track record of publishing work that challenges the status quo." Google declined to comment beyond Dean's statement.
Why it matters: Research from companies like Google forms a considerable portion of the broader body of work on key topics like AI. Gebru's case raises questions about the extent to which corporate research is shaped by corporate interests.
Our thought bubble: The conflict is in part a case of mismatched expectations.
- Companies should hardly be surprised if the diversity advocates and ethicists they hire point out problematic systems and behaviors inside their walls.
- At the same time, those researchers realistically can't expect the same level of academic freedom they would find working at an academic institution or nonprofit.
The bottom line: Cynthia Yeung, an industry veteran who spent five years at Google, put it bluntly: "Maybe the trade-off should be more clearly spelled out so researchers can make informed decisions before they accept a job offer: You get paid academic salaries in exchange for intellectual freedom, and you get paid Silicon Valley salaries in exchange for allowing your name/likeness to be used for brand/PR purposes and your research to be censored arbitrarily."
