Less than two years after Google fired two researchers who criticized biases built into AI systems, the company fired a researcher who questioned a paper it published about the capabilities of a specialized type of AI used to manufacture computer chips.
The researcher, Satrajit Chatterjee, led a team of scientists to challenge the celebrated research paper, which appeared last year in the scientific journal Nature, saying that computers could design certain parts of a computer chip faster and better than human beings. humans.
Dr. Chatterjee, 43, was fired in March, shortly after Google told his team it would not publish a paper refuting some of the claims made in Nature, four people familiar with the situation said. they were allowed to speak openly on the subject. matter. Google confirmed in a written statement that Dr. Chatterjee had been “terminated for cause.”
Google declined to elaborate on Dr. Chatterjee’s firing, but offered an outright defense of the research he criticized and his unwillingness to publish his assessment.
“We thoroughly reviewed the original Nature paper and endorse the peer-reviewed results,” Zoubin Ghahramani, vice president of Google Research, said in a written statement. “We also rigorously investigated the technical claims of a subsequent submission, and it did not meet our publication standards.”
Dr. Chatterjee’s firing was the latest example of discord in and around Google Brain, an AI research group seen as key to the company’s future. After spending billions of dollars to hire the best researchers and create new types of computing automation, Google has faced a wide variety of complaints about how it builds, uses and presents those technologies.
The tension between AI researchers at Google mirrors much larger struggles in the tech industry, which faces myriad questions about new AI technologies and the thorny social issues that have entangled these technologies and the people who build them.
Read more about Artificial Intelligence
The recent dispute also follows a familiar pattern of firings and whistleblowing among Google’s AI researchers, a growing concern for a company that has staked its future on infusing AI into everything it does. Sundar Pichai, CEO of Google’s parent company Alphabet, compared AI to the advent of electricity or fire, calling it one of humanity’s most important endeavors.
Google Brain started as a side project more than a decade ago when a group of researchers built a system that learned to recognize cats in YouTube videos. Google executives were so excited by the prospect of machines being able to learn skills on their own that they quickly expanded the lab, laying a foundation for remaking the company with this new artificial intelligence. The research group became a symbol of the company’s higher ambitions.
Before she was fired, Dr. Gebru was seeking permission to publish a research paper on how AI-based language systems, including technology created by Google, can end up using the biased and hateful language they learn from texts in books and websites. Dr. Gebru said that she had been exasperated by Google’s response to such complaints, including its refusal to publish the article.
A few months later, the company fired the other team leader, Margaret Mitchell, who publicly denounced Google’s handling of the situation with Dr. Gebru. The company said that Dr. Mitchell had violated its code of conduct.
The paper in Nature, published last June, promoted a technology called reinforcement learning, which the paper says could improve the design of computer chips. The technology was hailed as a breakthrough for artificial intelligence and a vast improvement on existing approaches to chip design. Google said it used this technique to develop its own chips for artificial intelligence computing.
Google had been working on applying the machine learning technique to chip design for years and published a similar paper a year earlier. Around that time, Google asked Dr. Chatterjee, who has a Ph.D. in computer science from the University of California, Berkeley, and had worked as a research scientist at Intel, to see if the approach could be sold or licensed to a design firm. chip. people familiar with the matter said.
But Dr. Chatterjee expressed reservations in an internal email about some of the paper’s claims and questioned whether the technology had been rigorously tested, three of the people said.
As the debate over that research continued, Google submitted another paper to Nature. For the presentation, Google made some adjustments to the previous article and removed the names of two authors, who had worked closely with Dr. Chatterjee and had also raised concerns about the article’s main claims, the people said.
When the most recent article was published, some Google researchers were surprised. They believed he had not followed a publication approval process that Jeff Dean, the company’s senior vice president who oversees most of its AI efforts, said was necessary after Dr. Gebru’s firing, the people said.
Google and one of the paper’s two lead authors, Anna Goldie, who wrote it with another computer scientist, Azalia Mirhoseini, said the changes to the previous paper didn’t require the full approval process. Google allowed Dr. Chatterjee and a handful of internal and external researchers to work on a paper that challenged some of his claims.
The team sent the rebuttal document to a so-called resolution committee for approval for publication. Months later, the article was rejected.
The researchers who worked on the rebuttal article said they wanted to take the issue to Mr. Pichai and Alphabet’s board of directors. They argued that Google’s decision not to publish the rebuttal violated its own AI principles, including advocating high standards of scientific excellence. Shortly after, Dr. Chatterjee was informed that he was no longer an employee, the people said.
Ms Goldie said Dr Chatterjee had asked to manage her project in 2019 and they had refused. When she later criticized him, she said, she failed to substantiate her claims and ignored the evidence they presented in response.
“Sat Chatterjee has waged a disinformation campaign against me and Azalia for over two years,” Ms Goldie said in a written statement.
She said that the work had been peer-reviewed by Nature, one of the most prestigious scientific publications. And she added that Google had used her methods to build new chips and that these chips were currently used in Google’s computer data centers.
Laurie M. Burgess, an attorney for Dr. Chatterjee, said it was disappointing that “certain authors of the Nature article are trying to shut down the scientific discussion by smearing and attacking Dr. Chatterjee simply for seeking scientific transparency.” Ms Burgess also questioned the leadership of Dr Dean, who was one of 20 co-authors on the Nature paper.
“Jeff Dean’s actions to suppress the publication of all relevant experimental data, not just data supporting his pet hypothesis, should be of deep concern to both the scientific community and the broader community that consumes Google products and services. Burgess said.
Dr. Dean did not respond to a request for comment.
After the rebuttal paper was shared with academics and other experts outside of Google, the controversy spread throughout the global community of researchers specializing in chip design.
Chipmaker Nvidia says it has used chip design methods that are similar to Google’s, but some experts aren’t sure what Google’s research means for the larger tech industry.
“If this really works well, that would be a big deal,” said Jens Lienig, a professor at the Dresden University of Technology in Germany, referring to the AI technology described in the Google article. “But it’s not clear if it’s working.”