Google faces internal battle over AI research to speed up chip design

Alphabet Inc’s Google said Monday that they had recently fired a senior chief engineer after colleagues whose landmark research into artificial intelligence software he had tried to discredit accused him of harassing behavior.

The dispute, which stems from efforts to automate chip design, threatens to undermine the reputation of Google’s research in academia. It could also disrupt the flow of millions of dollars in government grants for research into artificial intelligence and chips.

Google’s research unit has been under scrutiny since the end of 2020, after workers filed open criticism of its handling of staff complaints and publishing practices.

The new episode came after the scientific journal Nature in June published “A graph placement methodology for fast chip design,” led by Google researchers Azalia Mirhoseini and Anna Goldie. They discovered that AI could complete a key step in the chip design process, known as floor planning, faster and better than an unspecified human expert, a subjective reference point.

But other Google colleagues in a paper published anonymously online in March – "Stronger Baselines for Evaluating Deep Reinforcement Learning in Chip Placement" – found that two alternative approaches based on basic software surpass AI. One hit it on a familiar test, and the other on a proprietary Google box.

Google declined to comment on the leaked draft, but two workers confirmed its authenticity.

The company said it refused to publish Stronger Baselines because it did not live up to its standards, and shortly thereafter fired Satrajit Chatterjee, a leading driver of the work. It declined to say why it fired him.

“It’s unfortunate that Google has taken this turn,” said Laurie Burgess, a lawyer for Chatterjee. “It has always been his goal to have transparency about science, and he called on Google over the course of two years to address this.”

Google researcher Goldie told the New York Times, which on Monday first reported on the firing, that Chatterjee had harassed her and Mirhoseini for years by spreading misinformation about them.

Burgess dismissed the allegations, adding that Chatterjee did not leak Stronger Baselines.

Patrick Madden, an associate professor of chip design at Binghamton University who has read both articles, said he had never seen a paper before it in Nature that lacked a good point of comparison.

“It’s like a reference problem: Everyone gets the same puzzle pieces, and you can compare how close you get to getting everything right,” he said. “If they were to produce results on a standard benchmark and they were amazing, I would sing their praises.”

Google said the comparison with a human was more relevant and that software licensing issues had prevented it from mentioning tests.

Studies from major institutions like Google in well-known journals can have an overall impact on whether similar projects are funded in the industry. A Google researcher said the leaked paper had unfairly opened the door to questions about the credibility of any work published by the company.

Following “Stronger Baselines”; emerged online, Zoubin Ghahramani, vice president of Google Research, wrote on Twitter last month that “Google stands by this work published in Nature on ML for Chip Design, which has been independently replicated, open source and used in production at Google. “

Nature, referring to a British holiday, had no immediate comment. Madden said he hoped Nature would revisit the publication, noting that peer reviewer notes show that at least one asked for results on benchmarks.

Somehow it never happened," he said.

Leave a Reply

Your email address will not be published.