On February 19, Google fired Margaret Mitchell, the AI ethics co-lead at Google Brain. Mitchell’s co-leading colleague, Timnit Gebru, had been fired in December, amid controversy. Both women were critical of Google’s diversity hiring record during the two years they worked together. The flashpoint in Mitchell’s case, for which she had been temporarily suspended earlier, hinged on claims of unauthorized use of files:
In a statement, a Google spokesperson said Mitchell had shared “confidential business-sensitive documents and private data of other employees” outside the company. After Mitchell’s suspension last month, Google said activity in her account had triggered a security system. A source familiar with Mitchell’s suspension said she had been using a script to search her email for material related to Gebru’s time at the company.Tom Simonite, “A Second AI Researcher Says She Was Fired by Google” at Wired
She had also criticized company management via Twitter for insufficient diversity in hiring: “Say you have a problem with consistently alienating Black women and have caused serious damage in their lives. You could: A) try to undo that damage B) try to find more Black people to like you (the tokenism approach). Good luck…..” Mitchell also wrote a public letter protesting her colleague Gebru’s firing. About 2600 Googlers have also protested Gebru’s departure in a letter reproduced at Medium.
Gizmodo has more details about the flashpoint issue in Mitchell’s case:
Mitchell, who formerly led the team alongside Gebru, was caught using automated scripts to comb through her work emails to find evidence of discrimination and harassment to back up Gebru’s claims, Axios reports. In January, she lost access to her corporate email after Google launched an investigation into her activity. In a statement to Reuters, Google claims Mitchell’s firing followed disciplinary recommendations by investigators and a review committee. Google said she violated the company’s code of conduct and security policies and transferred electronic files outside the company.Alyse Stanley, “Google Fires Another Top Researcher on Its AI Ethics Team” at Gizmodo
Timnit Gebru had balked at removing her name from an academic paper that criticized large language models in search engines (including one that made Google a good deal of money) as insensitive to diversity issues, among other failings.
Google claims that the paper fell short of its standards. However, it has since been accepted and will be presented at a virtual conference in March, FAccT ’21, March 3–10, 2021, Virtual Event, Canada. A key focus of the conference will be “scholarship on systemic racism and discrimination, use of AI in health and health inequities, and the growing demand for new forms of accountability around the world,” according to ACM FAccT General Co-chair William Isaac, a senior research scientist with DeepMind, quoted in a conference media release.
Gebru and Mitchell could soon be joined in the job search by a third Google AI ethics researcher, Alex Hanna:
Another leading AI ethics researcher at the company, Alex Hanna, tweeted a screenshot that evening purportedly of an email she received from higher-ups threatening that she is next on the chopping block.Alyse Stanley, “Google Fires Another Top Researcher on Its AI Ethics Team” at Gizmodo
Google has since appointed Vice President of Engineering Marian Croak to fill the gaps. The appointment has been met with criticism from Gebru: “Meanwhile, Gebru reacted strongly against the appointment and expressed her displeasure at Google’s alleged attempt at pitting one black woman against another.” (Analytics India, February 22, 2021). Meanwhile, Google is attempting various fixes to stem the tide of criticism over its handling of the issues.
It may be a bit early for indepth analysis of what it all means but Tom Simonite observes at Wired,
The two women’s acrimonious exits from Google have drawn new attention to the tensions inherent in companies seeking profits from AI while also retaining staff to investigate what limits should be placed on the technology. After Gebru’s departure, some AI experts said they were unsure whether to trust Google’s work on such questions.Tom Simonite, “A Second AI Researcher Says She Was Fired by Google” at Wired
Reuters offers another take:
Google has recruited top scientists with promises of research freedom, but the limits are tested as researchers increasingly write about the negative effects of technology and offer unflattering perspectives on their employer’s products.
Reuters reported exclusively in December that Google introduced a new “sensitive topics” review last year to ensure that papers on topics such as the oil industry and content recommendation systems would not get the company into legal or regulatory trouble. Mitchell publicly expressed concern that the policy could lead to censorship.Paresh Dave, Jeffrey Dastin, “Google fires second AI ethics leader as dispute over research, diversity grows” at Reuters (February 19, 2021)
In the end, the AI ethics team and Google management may have very different ideas about what “ethics” means. To the AI ethics team, promotion of equity and diversity has been a top ethical issue; Google management may see that more as a highly desirable thing to pursue but not at the expense of profits or company policy with respect to document security.
For now, all we can be sure of is, we ain’t heard the last of it.
You may also wish to read:
Google’s leading AI ethics researcher fired, amid controversy. Her research team targeted Google’s “cash cow”: advertising. Timnit Gebru joins a number of ex-Googlers forced out by such issues in recent years. Her paper, while troubling, should have been given an honest hearing. (December 8, 2020)
Google’ secret health data grab: The whistleblower talks: This is the fourth whistleblower in the last eighteen months. (November 15, 2019)