Less than two years after Google fired two researchers for criticizing the bias built into artificial intelligence systems, the company questioned a paper it published about the capabilities of a special type of artificial intelligence used to make computer chips. The researcher was fired.
Researcher Satrajit Chatterjee led a team of scientists on a famous challenge. research paperIn a paper last year published in the scientific journal Nature, he said that computers can design certain parts of computer chips faster and better than humans.
Chatterjee, 43, was fired shortly after Google told his team that it would not publish papers that refute some of Nature’s claims. Problem. In a written statement, Google confirmed that Dr. Chatterjee was “fired for cause.”
Google declined to elaborate on Dr. Chatterjee’s dismissal, but offered a full-blown defense of the study he criticized and his reluctance to publish his assessment.
“We have thoroughly researched the original Nature paper and support the peer-reviewed findings,” said Zoubin Ghahramani, vice president of Google Research, in a written statement. “We also rigorously scrutinized the technical claims in subsequent submissions, but they did not meet the criteria for publication.”
Dr. Chatterjee’s layoff was the most recent example of a feud within and around Google Brain, an AI research group that is believed to be the key to the company’s future. After hiring the best researchers and spending billions of dollars creating new kinds of computer automation, Google has grappled with a variety of complaints about how these technologies are built, used, and portrayed.
Tensions among AI researchers at Google reflect an even greater struggle across the tech industry, facing countless questions about new AI technologies and the tangible societal issues entwining them and the people who build them.
Learn more about artificial intelligence
The recent dispute also follows a false pattern of dismissal and duel allegations among AI researchers at Google. This is a growing concern for companies betting their future on injecting artificial intelligence into everything. Sundar Pichai, CEO of Google’s parent company Alphabet, likens artificial intelligence to the arrival of electricity or fire, calling it one of mankind’s most important efforts.
Google Brain started as a side project more than a decade ago when a group of researchers built a system that learned how to recognize a cat from a YouTube video. Driven by the expectation that machines could learn technology on their own, Google executives quickly expanded their lab, paving the way for rebuilding the company with this new artificial intelligence. The research group has become a symbol of the company’s greatest ambitions.
Before she was fired, Dr. Gebru sought permission to publish research papers on how AI-powered language systems, including those built by Google, might end up using biased and aversive language learned from text in books and websites. I did. Dr. Gebru said she was increasingly annoyed by Google’s response to these complaints, which included Google’s rejection of the paper.
A few months later, the company fired another team leader, Margaret Mitchell. Margaret Mitchell publicly criticized Google for handling the situation with Dr. Gebru. The company said Dr. Mitchell violated her code of conduct.
A paper published in Nature last June promoted a technique called reinforcement learning, which said it could improve the design of computer chips. The technology has been hailed as a breakthrough for artificial intelligence and a significant improvement over existing approaches to chip design. Google said it used the technology to develop its own chips for artificial intelligence computing.
Google has been working on applying machine learning techniques to chip design for many years. similar paper 1 year ago. Around that time, Google asked Dr. Chatterjee, a researcher at Intel with a PhD in computer science from the University of California, Berkeley, to see if the approach could be sold or licensed to a chip design company. , said people versed in the matter.
However, Dr. Chatterjee took note of some of the paper’s claims in an internal email and asked whether the technology had been rigorously tested, three people said.
While the debate over the study continues, Google has published another paper in Nature. For submission, people said Google made minor revisions to the previous paper, worked closely with Dr. Chatterjee, and removed the names of two authors who expressed concern about the paper’s main claims, people said.
When a new paper was published, some Google researchers were surprised. They didn’t follow the publication approval process that Jeff Dean, the company’s senior vice president, who oversees most of its AI efforts, said was necessary in the wake of Dr. Gebru’s dismissal, people said.
Anna Goldie, one of two lead authors on the report, along with Google and fellow computer scientist Azalia Mirhoseini, said the changes in the previous paper did not require a full approval process. Google has allowed Dr. Chatterjee and a handful of internal and external researchers to write papers that challenge some of the claims.
The team submitted the rebuttal paper to a so-called resolution committee for publication approval. A few months later, the paper was rejected.
The researchers who studied the rebuttal paper said they wanted to report the matter to Pichai and Alphabet’s board of directors. They argued that Google’s decision not to post a counter-argument violated itself. AI Principles, including maintaining a high level of scientific excellence. Shortly thereafter, people said Dr. Chatterjee was informed that he was no longer an employee.
Goldie said that Dr. Chatterjee asked to manage the project in 2019 but turned it down. When he later criticized it, she said she was unable to substantiate her own complaints and ignored the evidence they presented in her response.
“Sat Chatterjee has been campaigning for more than two years with misinformation about me and Azalia,” Goldie said in a written statement.
She said the study was peer-reviewed in one of the most prestigious scientific publications, Nature. And she added that Google used their method to make the new chip, which is now being used in Google’s computer data centers.
Dr. Chatterjee’s attorney, Laurie M. Burgess, said: “Certain authors of the Nature papers seek to halt scientific discussion by defaming and attacking Dr. Chatterjee simply for seeking scientific transparency.” Burgess is also one of 20 co-authors of the Nature paper, Dr. Dean’s leadership was questioned.
“Jeff Dean’s behavior to curb the release of all relevant experimental data, as well as data to support his preferred hypothesis, will be of deep concern to both the scientific community and the broader community that consumes Google’s services and products,” Burgess said. said.
Dr. Dean did not respond to requests for comment.
After the rebuttal papers were shared with academia and other experts outside of Google, the controversy spread to the global community of researchers specializing in chip design.
Chip maker Nvidia said it used a chip design approach similar to Google, but some experts aren’t sure what Google’s research will mean for the large tech industry.
“If this works really well, it would be really great,” said Jens Lienig, a professor at the Dresden University of Technology in Germany, referring to the AI technology described in a Google paper. “But it’s unclear whether that would work.”
Another launch between Google’s AI Brain Trust, and more feuds
Source link Another launch between Google’s AI Brain Trust, and more feuds