Google Faces Internal Dispute in AI Research for Chip Development
The episode emerged after the scientific journal Nature published in June a paper led by Google scientists Azalia Mirhoseini and Anna Goldie, suggesting that AI could complete a key step in the chip design process faster and better than a human expert.
However, other Google colleagues, in an anonymous article posted in March, showed that alternative approaches based on basic software outperformed the AI. One surpassed it in a well-known benchmark test, and the other in a proprietary Google benchmark.
Google declined to comment on the leaked draft, but two employees confirmed its authenticity.
The company said it chose not to publish the second article because it didn’t meet its standards, and shortly afterward, it fired Satrajit Chatterjee, one of the main proponents of the paper. Google declined to explain the reason for the dismissal.
“It’s a shame that Google took this path,” said Laurie Burgess, Chatterjee’s attorney. “It was always his goal to ensure transparency in science, and he asked for two years for Google to resolve this.”
Researcher Anna Goldie told The New York Times that Chatterjee had spread false information about her and Mirhoseini for years. Burgess denied the allegations.
Studies from major institutions like Google, when published in renowned journals, can have a significant impact on funding in the industry.
A Google researcher said the leaked article unfairly opened the door to questions about the credibility of any work published by the company.
Source: CNN
Follow our profile on social media and stay up to date!
