2 min

Tags in this article

, ,

Google announced on Monday that a senior engineering manager had been fired after colleagues, whose groundbreaking artificial intelligence research he was attempting to discredit, accused him of harassment. 

The controversy, which arises from Google’s efforts to automate semiconductor design, threatens the unit’s academic standing. It also can hinder the flow of millions of dollars in government funding for AI and semiconductor development.

Since late 2020, Google’s research division has been under fire as employees expressed open dissatisfaction with the company’s handling of personnel grievances and publishing policies.

What happened?

The current episode began when Google scientists Azalia Mirhoseini and Anna Goldie published “A graph placement methodology for fast chip design” in the scientific journal Nature in June.

They observed that AI could perform a critical phase in the semiconductor design process known as floorplanning faster and better than an unnamed human expert, a subjective benchmark.

However, other Google employees discovered that two alternative ways based on fundamental software beat AI in a document published anonymously online in March titled “Stronger Baselines for Evaluating Deep Reinforcement Learning in Chip Placement.”

One topped it on a well-known test, while the other beat it on a Google-created criterion.

It’s a whole thing

Google declined to comment on the document, but two employees at the company verified its validity.

Stronger Baselines, Google said, was not published because it did not match corporate requirements, and Satrajit Chatterjee, a key driver of the project, was sacked shortly after. Google refused to reveal why he was sacked.

Goldie, a Google researcher, told the New York Times on Monday that Chatterjee had been harassing her and Mirhoseini for years by disseminating false information about them.

Burgess refuted the claims, claiming that Chatterjee did not release Stronger Baselines. How this affects Google, only time will tell. The corporation continues to do all the wrong things regarding how it treats its AI researchers. 

Also read: Google AI makes webshops more personalized.