Tension inside Google over conduct of fired AI researcher

Asked about Chatterjee, Google spokesman Jason Freidenfelds provided a statement from the company confirming he was “terminated with cause.” Freidenfelds also provided a statement from Zoubin Ghahramani, vice president of Google Research, saying that “we firmly maintain our standard of respectful speech among our researchers.” Gharamani’s statement did not mention Chatterjee by name.

The episode adds to a series of recent internal conflicts at Google that hint at the freewheeling, engineer-centric culture it celebrated as a startup. left the company unprepared for some of the challenges of being a multinational with over 100,000 employees.

Google hired Satrajit Chatterjee in 2018 as a senior machine learning researcher. He was previously senior vice president of the hedge fund Two Sigma and had also worked at Intel. When Chatterjee joined the team, Mirhoseini and Goldie were already working in the company’s largest machine learning lab, Google Brain. Chatterjee joined a separate, smaller research group within Google’s search division.

Advertising

The two women did not work directly with Chatterjee, but in 2019, according to Goldie’s internal document, he asked to manage the Morpheus project. After being politely declined, employees say, Chatterjee began raising doubts about the duo’s work with senior researchers whom they needed to collaborate with or gain support, suggesting their work was fake or even fabricated.

As a more senior employee, Chatterjee’s questions could be influential. As a result, employees say, other senior managers have sometimes become skeptical of Goldie and Miroseini’s work, questioning their results.

The effect has been to turn Miroseini and Goldie’s job at Google into a stressful and divisive reality, insiders claim. Along with running a successful project with support from Google’s chip designers, they say the duo had to do extra work to address claims that their results were bogus or even bogus.

Chip design teams at Google and elsewhere are generally conservative by nature because nanoscale manufacturing is expensive and errors in a chip cannot be corrected once it’s been cut from silicon. Google said the TPUs have enabled breakthroughs in its AI research and services, and leases the chips through its cloud unit. Yet Chatterjee’s criticism of Morpheus continued even after Google’s hardware executives decided they trusted him enough to let him help design the company’s next generation of TPUs.

In May 2021, a Google employee posted to an internal mailing list asking if anyone had applied machine learning to circuit board design. Mirhoseini responded by saying that Morpheus could help. But Chatterjee stepped in to say that older techniques outperformed machine learning tools and that commercially available chip design tools provided the best results.

Jeff Dean, head of AI at Google, joined the discussion to say that Morpheus was already being used to design the next generation of TPU chips. The technology won extensive tests against human chip experts and commercial chip design tools, Dean said, while attaching a slideshow of the results.

Dean also linked to the team’s recently published, peer-reviewed report. Nature to study. He reported that the Morpheus team’s code presented TPU circuit blocks better than Google engineers using commercial chip design tools. The authors did not disclose details of these chip segments, saying they were confidential to Google, but also included results for an open-source processor design that is freely available to everyone. The results of the article were then replicated by another research team within Google, and the code for the experiments was open sourced.

Leave a Comment