Alphabet Inc. Google announced Monday that it had recently fired a senior engineering director after colleagues whose artificial intelligence software landmark research was trying to discredit him accused him of harassing him. The controversy, stemming from chip design automation efforts, threatens to undermine 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 unit has been under scrutiny since late 2020, with employees openly criticizing staff complaints handling and publishing practices. The new episode came after the science journal Nature in June published the “Graphics Placement Methodology for Quick Chip Design”, led by Google scientists Azalia Mirhoseini and Anna Goldie. They discovered that artificial intelligence could complete a key step in the chip design process, known as floor design, faster and better than an unspecified human expert, a subjective benchmark. But other Google colleagues found in a paper anonymously posted online in March – “Stronger Baselines for Evaluating Deep Reinforcement Learning in Chip Placement” – that two key software-based alternatives outweighed artificial intelligence. One won it in a well-known test and the other in an exclusive Google rubric. Google declined to comment on the leaked draft, but two employees confirmed its authenticity. The company said it refused to publish Stronger Baselines because it did not meet its standards and immediately fired Satrajit Chatterjee, a leading project guide. He refused to say why he was fired. “It’s unfortunate that Google has taken this turn,” said Laurie Burgess, a Chatterjee lawyer. “It has always been his goal to be transparent about science and he has been urged by Google for two years to address it.” Google researcher Goldie told the New York Times, which on Monday first reported the dismissal, that Chatterjee had been harassing her and Mirhoseini for years by spreading misinformation about them. Burgess denied the allegations, adding that Chatterjee had not leaked the Stronger Baselines. Patrick Madden, an associate professor who focused on chip design at Binghamton University and who has read both papers, said he had never seen a work before that in Nature that did not have a good point of comparison. “It’s like a reference problem: Everyone gets the same puzzle pieces and you can compare how close you got to doing everything right,” he said. “If they were working at a typical benchmark and they were great, I would praise them.” Google said the comparison to a human was more relevant and that software licensing issues prevented it from reporting tests. Studies by major institutions such as Google in well-known journals can have a major impact on whether such projects are funded in the industry. A Google researcher said the leaked document unjustly opened the door to questions about the credibility of any work published by the company. Following the release of “Stronger Baselines” on the Internet, Zoubin Ghahramani, vice president of Google Research, wrote on Twitter last month that “Google supports this work published in Nature in ML for Chip Design, which has been reproduced independently, open source and used in production by Google. “ Nature, citing a UK holiday, did not comment immediately. Madden said he hoped Nature would review the publication, noting that peer reviewers’ notes show that at least one requested results on the benchmarks. “In a way, that never happened,” he said. (Report by Paresh Dave. Edited by Gerry Doyle)
