Tension Inside Google Over a Fired AI Researcher’s Conduct

Rate this post

When asked about Chatterjee, Google spokesman Jason Freidenfelds provided a statement from the company confirming that it was “terminated for cause”. Freidenfelds also made a statement from Zoubin Ghahramani, vice president of Google Research, saying that “we firmly hold to our standard for respectful discourse 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 suggest that the free, engineering-centric culture it held as a startup has left the company unprepared for the challenges of being a multinational with more than 100,000 employees.

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

The two women did not work directly with Chatterjee, but in 2019, according to Goldie’s internal document, she asked to manage the Morpheus project. After being politely rejected, employees say, Chatterjee began questioning the couple’s work with senior researchers with whom they needed to collaborate or gain support, suggesting their work was wrong or even fabricated.

As a senior employee, Chatterjee’s questions could be influential. As a result, employees say, other senior staff members became skeptical of Goldie and Miroseini’s work, questioning their findings.

The effect was to turn Miroseini and Goldie’s work on Google into a stressful and divided reality, experts say. While running a successful project with the support of Google’s chip designers, they say the couple had to do extra work to respond to allegations that their results were incorrect or even false.

Chip design equipment from Google and elsewhere is often cautious by nature, because nanoscale manufacturing is expensive and errors in a chip cannot be fixed once it has been cut into silicon. Google has said that TPUs have made progress in its research and AI services, and it is renting out chips through its cloud drive. Still, Chatterjee’s criticism of Morpheus continued even after Google’s hardware leaders decided they trusted him enough to let him help design the company’s next-generation TPU.

In May 2021, a Google employee posted to an internal email list asking if anyone had applied machine learning to circuit board design. Mirhoseini responded by saying that Morpheus could help. But Chatterjee intervened to claim 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 by saying that Morpheus was already being used to design the next generation of TPU chips. The technology had won extensive testing against human chip experts and commercial chip design tools, Dean said, while attaching a platform to the results slide.

Dean has also been linked to the team’s recently reviewed peer-reviewed post Nature to study. He reported that Morpheus ’team code set TPU circuit blocks better than Google engineers who used 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 a free open source processor design available to anyone. The results of the document were later replicated by another search team within Google, and the code of the experiments was open source.

Source link

Leave a Comment