He bet on the award, he said, “because of my feeling at the time that Google was building a very strong, potentially industry-leading ethical AI team.”
At the time, Gebru said Google AI management asked her to withdraw the paper from review for presentation at a conference, or remove her name from it. Google said it accepted Gebru’s resignation because of a list of requirements she had sent via email that had to be met in order for her to continue working for the company.
All of this came into sharp focus for Stark on Wednesday, March 10, when Google sent him a congratulatory note giving him $ 60,000 for his proposal for a research project that would look at how companies roll out AI used to detect emotions. Stark said he immediately felt he needed to turn down the award in order to show his support for Gebru and Mitchell, as well as those still on the ethical AI team at Google.
“My first thought was, ‘I have to reject it,'” Stark told CNN Business.
“With a clear conscience, I can no longer accept funding from a company that treats its employees this way,” Vijay Chidambaram, an assistant professor at the University of Texas at Austin who studies storage systems, told CNN Business. Chidambaram previously received $ 30,000 from Google in 2018 for a research project.
The money involved is of little importance to Google. But the growing fallout from Google’s tensions with the ethical AI team now poses a risk to the company’s reputation and growth in the AI community. This is crucial as Google is fighting for talent – both as employees of the company and names associated with it in the academic community.
“I think this is more widespread than even the company is aware of,” Stark said.
Decline in solidarity
Despite his initial inclination, Stark did not immediately deny Google’s award. He talked to colleagues about what he planned to do – “People supported what decision I made,” he said – before sending Google his answer the following Friday. He thanked the company for the “vote of confidence” in the research, but wrote that he “rejected this award in solidarity with Dr. Gebru and Mitchell, their teammates and all those who have been in similar situations”, according to an email shown by CNN Business.
“I look forward to working with Google Research again as soon as the organization and its leaders have reflected on their decision in this matter, addressed the damage they have caused, and committed, in word and deed, to promote critical research and products. which supports justice, ”Stark wrote.
Gebru said she appreciated Stark’s action.
“It’s a pretty big deal for someone to reject Google sponsorship,” she told CNN Business. “Especially someone who is early in their career.”
A Google spokesman said that in the last 15 years, the company has provided over 6,500 academic and research grants to those outside Google. Stark is the first person to say no, according to the spokesperson.
“It was a real failure as they were treated”
Still, Stark’s decision is just the latest show of solidarity with Gebru and Mitchell.
“It was a real failure [Gebru and Mitchell] was treated. No one apologized to them yet, “she told CNN Business in a recent interview. I do not want to interact with companies that behave this way towards top researchers. ”
Google’s efforts to push boundaries in AI
Google is aware that its reputation as a research institution has been damaged in recent months, and the company has said it intends to fix it. In a recent town hall meeting on Google, which Reuters first reported on and CNN Business also received word from, the company outlined changes it makes to its internal research and publication practices.
“I think the way to regain trust is to continue to publish groundbreaking work in many, many areas, including pushing the boundaries of responsible AI-related topics, publishing things that are deeply interesting to the research community, I think is one of the best ways to continue to be a leader in the field of research, “said Jeff Dean, Google’s Head of AI. He answered an employee’s question regarding external researchers and said they would read articles from Google “with more skepticism now.”
Gebru hopes that, like the FAccT, more conferences will reassess their relationship with the technology companies’ research laboratories. Historically, much of the work of developing and studying AI has been done in an academic setting. But as companies have found more and more commercial use for the technology, the lines between academia have become blurred. Google is just one of many technology companies that have great influence over academic conferences that publish many of the researchers’ articles; Its employees sit on conference boards, and it sponsors many conferences each year, sometimes for tens of thousands of dollars.
For example, Google and some subsidiaries of the parent company Alphabet were listed as sponsors at the $ 20,000 “platinum” and $ 10,000 “gold” level at the International Conference on Machine Learning, or ICML, and the Conference on Neural Information Processing Systems, or NeurIPS, in 2020 – both major AI conferences. And some of the company’s employees sit on the organizing committees.
ICML President John Langford said the conference is “currently open for sponsorship” by Google for the July 2021 conference.
“There is a lot of discussion going on about how ICML as a conference should encourage a good machine learning culture and practice with future sponsorship policies as part of the discussion,” he added.
NeurIPS CEO Mary Ellen Perry said the conference has not yet made its annual call for sponsorship, but that requests “will be considered against a set of selection guidelines in place by this year’s sponsor chairs”; NeurIPS is scheduled for December.
For Stark and others in the academic research community, however, their criteria for accepting funding from Google has already changed.
“Extra research money would be good,” Stark said. “But there was something I felt I just could not take.”