Google calls AI scientists ‘strike a positive tone’ in research


Illustration for an article titled Google Reportly Attacking AI Scientists in a Positive Tone in Research

Photo: Leon Neil / Staff (Getty Images)

About three weeks after Blake’s sudden exit Artificial intelligence ethics Timnit Gabru, Google More details are coming out about the shady new set of policies designed for its research team.

After reviewing internal communications and talking to researchers affected by the rule change, Reuters The tech giant reported Wednesday that it has recently added a “sensitive subject” review process to its scientists’ papers, and on at least three occasions has made a clear request that scientists refrain from casting Google’s technology in a negative light.

Under the new procedure, scientists need to meet with specialized legal, policy and public relations teams before proceeding with AI research on so-called controversial issues, which may include facial analysis and classification of gender, gender or political affiliation.

In an example reviewed by Reuters, scientists who recommended that the AIA use it to create user feeds on platforms such as YouTube – a property owned by Google – prepared a paper in which it is concerned that the use of this technology is “disinformation, discriminatory.” Or otherwise to encourage. Inappropriate results “and” insufficient diversity of content “, as well as” political polarization “. After a review by a senior manager who instructed researchers to strike a more positive tone and the final release suggests that the systems “can promote accurate information, fairness and diversity of content.”

“Advances in technological ad g and the increasing complexity of our external environment are increasingly leading to situations where potentially inadvisable projects raise ethical, reputable, regulatory or legal issues,” said an internal webpage outlining the policy. .

In recent weeks – and especially then Departure of GabruAfter raising alarms about censorship infiltrating the research process, a well-known researcher reportedly favored higher ups – Google has faced further scrutiny of potential biases in its internal research department.

Four staff researchers who spoke to Reuters acknowledged Gabru’s claims, saying they also believe Google is starting to interfere in a critical study of the potential for harm.

Margaret Mitchell, “If we’re researching the right thing to do with our expertise, and we’re not allowed to publish that is not consistent with a high quality peer review, then we’re in serious trouble with censorship,” said Margaret Mitchell, one of the company’s Senior Scientist Dr.

In early December, Gabru claimed he had been fired by Google. She claimed after pushing against the order not to publish the research that an AI capable of mimicking speech could put the marginalized population at bay.

.