Google will change the procedures for reviewing scientists’ work before July, according to a City Hall poll by Reuters as part of an effort to quell internal unrest over the integrity of its artificial intelligence (AI).
In comments at a staff meeting last Friday, Google Research executives said they were regaining trust after the company expelled two prominent women and rejected their jobs, according to a one-hour survey whose content was confirmed by two sources. is.
Maggie Johnson, chief operating officer of the research unit, said at the meeting that teams are already testing a questionnaire that will assess projects on risks and help scientists navigate the reviews. This initial change would begin by the end of the second quarter, and most newspapers do not need additional investigation, she said.
Reuters reported in December that Google had instituted a “sensitive topic” review for studies on dozens of issues, such as China or bias in its services. Internal reviewers have demanded that at least three articles on AI be amended so as not to negatively impact Google technology, Reuters reports.
Jeff Dean, senior vice president of Google, which oversees the division, said on Friday that the review on “sensitive topics” was confusing and that he had instructed a senior research director, Zoubin Ghahramani, to follow the rules according to the survey. to clear up. .
Ghahramani, a Cambridge University professor who joined Google from Uber in September, said during City Hall: “We need to feel comfortable with the discomfort” of self-critical research.
Google declined to comment on Friday’s meeting.
An internal email, seen by Reuters, provides fresh details about the researchers’ concerns about Google and shows exactly how Google’s legal department amended one of the three AI articles, called Extracting Data from Large Language Models.
The email, dated Feb. 8, from a co-author of the newspaper, Nicholas Carlini, went to hundreds of colleagues to draw their attention to what he calls “deeply treacherous” amendments by company lawyers.
“Let’s be clear here,” said the 1,200-word email. “If we as academics write that we have a ‘concern’ or find something ‘worrying’ and a Google lawyer requires us to change it to sound nicer, it’s very big that Big Brother will step in.”
According to his email, the required amendments included ‘negative-to-neutral’ barter transactions, such as changing the word ‘concern’ to ‘considerations’ and ‘dangers’ to ‘risks’. Lawyers also had to remove references to Google technology; the findings of the authors that AI leaked copyrighted content; and the words “offense” and “sensitive”, read the email.
Carlini did not respond to requests for comment. Google, in response to questions about the email, disputes its claim that lawyers are trying to control the tone of the newspaper. The company said it had no problems with the topics investigated by the newspaper, but it found that some legal terms were used inaccurately and consequently amended them thoroughly.
Racial Equity Audit
Last week, Google also appointed Marian Croak, a pioneer in Internet audio technology and one of Google’s few black vice presidents, to consolidate and manage ten teams to study issues such as racial bias in algorithms and technology for people with disabilities.
Croak said at Friday’s meeting that it would take time to address concerns among AI ethics researchers and mitigate damage to Google’s brand. “Please hold me fully responsible for trying to reverse the situation,” she said on the recording.
Johnson added that the AI organization is introducing a consulting firm for a wide-ranging impact on racial equity. The first of its kind audit for the department will lead to recommendations “that are going to be quite difficult”, she said.
Tensions in Dean’s division deepened in December after Google fired Timnit Gebru, co-leader of its ethical AI research team, after her refusal to withdraw a paper on language-generating AI. Gebru, who is Black, at the time accused the company of judging her work differently because of her identity and of marginalizing employees from under-represented backgrounds. Nearly 2,700 employees signed an open letter in support of Gebru.
During City Hall, Dean elaborated on which scholarship the company would support. “We want responsible AI and ethical AI research,” Dean said, citing the study of the environmental costs of technology. But he said it was problematic to quote the data ‘almost a factor of a hundred’, while ignoring more accurate statistics as well as Google’s efforts to reduce emissions. Dean had earlier criticized Gebru’s article for not containing important findings on the environmental impact.
Gebru defends the quote from her paper. “It’s really bad for Google to put it defensively against an article quoted by so many of their peer-to-peer institutions,” she told Reuters.
Employees continued to post on Twitter over the past month about their frustrations when Google investigated Margaret Mitchell, an ethical co-leader, and then fired her for moving electronic files outside the company. Mitchell said on Twitter that she had acted “to raise concerns about racial and gender inequality, and to talk about Google’s problematic dismissal of Dr Gebru”.
Mitchell collaborated on the newspaper that led to the departure of Gebru, and a version published online last month without Google commitment, named “Shmargaret Shmitchell” as co-author.
Asked for comment, Mitchell was disappointed by a lawyer in Dean’s criticism of the newspaper, saying her name had been removed following a company order.