Pichai apologised to staff who may “question their place” after her departure, which raised issues over workforce diversity and the ethics of AI. Dr Gebru asked for an explanation for why she had to withdraw a paper highlighting flaws in language technology, and when none was given, she offered to resign after a short period. Google accepted it immediately, while 2,000 employees signed a petition protesting her leaving.
Why does this matter?
Google, alongside other big Silicon Valley firms, is currently facing regulation and backlash for its monopolistic practices. While the lawsuits filed specifically refer to Google’s strategies as a search engine, when looking at the influence of Big Tech more widely these firms also have a tight hold on technological innovation at large – especially when looking at AI and machine learning development.
Over recent years we’ve seen Google create various opportunities to fuel and protect the development of its AI technology. This has included acquiring small innovative firms (a process that typically involves refocusing development according to Google’s own targets) and developing its own educational qualification – Google Career Certificates. This firm has also been a front-runner in discussions about regulations for controversial technology such as facial-recognition.
What we should note about all of these moves is that not only do they bolster the internal capabilities of Google, they also set the direction, methodology and (self-)regulation for AI development for the future. When it comes to setting standards around AI, all roads lead to Google and other Big Tech firms.
Against this backdrop, with Google able to dictate how its AI is being developed, what it’s being used for, and what regulations it abides by, losing employees who offer research and perspectives challenging its strategies potentially means losing the last set of checks and balances to its work and the last line of defence for consumers.
Lateral thought from Curation
Seemingly aware of the issue of inherent bias filtering into technological development, the incoming Biden-Harris Administration has previously talked about increasing diversity within the industry. “Diversifying” algorithms, however, may not be as simple as changing the face of the workforce.
It’s been suggested, for example, that “fixing” biased algorithms could even violate US discrimination laws, albeit unintentionally. Such laws only govern for intent and not outcome, meaning intentionally altering technology could be seen as biased in its own right.