Sign up now for free and unlimited access to Reuters.com
register
OAKLAND, Calif., May 2 (Reuters) – Alphabet Inc’s (GOOGL.O) Google said on Monday it recently fired a senior engineering manager after trying to discredit his landmark research into artificial intelligence software colleagues accused him of harassment.
The dispute, which stems from efforts to automate chip design, threatens to damage Google’s research reputation in academia. It could also disrupt the flow of millions of dollars in government grants for artificial intelligence and chip research.
Google’s research division has faced scrutiny since late 2020 after workers publicly criticized its handling of personnel complaints and posting practices.read more
Sign up now for free and unlimited access to Reuters.com
register
The new episode comes after the June issue of the scientific journal Nature published “A Pattern Placement Approach for Rapid Chip Design,” led by Google scientists Azalia Mirhoseini and Anna Goldie. They found that AI can complete a key step in the chip design process, floor planning, faster and better than unspecified human experts (subjective reference points).
But other Google colleagues found in a paper published anonymously online in March that two alternatives based on basic software outperformed artificial intelligence. One beat it on the famous test, the other beat it on the proprietary Google rules.
Google declined to comment on the leaked draft, but two staff members confirmed its authenticity.
The company said it refused to release the Stronger Baseline because it did not meet its standards, and fired Satrajit Chatterjee, a key driver of the effort, shortly after. It declined to say why it fired him.
“It’s unfortunate that Google has taken that stance,” said Chatterjee’s attorney Laurie Burgess. “Keeping the science transparent has always been his goal, and he’s been urging Google to address this for two years. question.”
Goldie, a Google researcher, told The New York Times, which first reported the firing on Monday, that Chatterjee had been harassing her and Mir Hosseini for years by spreading misinformation about them.
Burgess has denied the allegations, adding that Chatterjee did not divulge stronger baselines.
Patrick Madden, an associate professor of chip design at Binghamton University who has read the two papers, said he had never seen one before the Nature paper that lacked a good point of comparison. a dissertation.
“It’s like a reference problem: everyone gets the same puzzle pieces, and you can compare how well you got everything right,” he said. “If they’re going to produce results on some standard benchmark and they’re good, I’ll give them credit for that.”
Google said the human comparison was more relevant, and software licensing issues prevented it from mentioning the test.
Research in reputable journals by large institutions such as Google can have a huge impact on whether similar projects in the industry are funded. A Google researcher says the leaked paper unfairly opens the door to questioning the credibility of any work the company publishes.
After a “stronger baseline” emerged online, Google Research VP Zoubin Ghahramani tweeted last month: “Google supports this work published in Nature on ML for Chip Design, which has been independently reproduced , open source, and used in production at Google.”
Nature, citing the UK public holiday, had no immediate comment. Madden said he hoped Nature would revisit the publication, noting that peer reviewer notes showed at least one request for benchmark results.
“Somehow this never happened,” he said.
Sign up now for free and unlimited access to Reuters.com
register
Paresh Dave reports.Edited by Gerry Doyle
Our Standard: The Thomson Reuters Trust Principles.