Kelsey Piper* says ethics boards like the one just announced by Google don’t appear to be equipped to solve, or even make progress on, hard questions about the progress of AI.
Photo: Paweł Czerwiński
Just a week after it was announced, Google’s new AI ethics board was already in trouble.
The board, founded to guide “responsible development of AI” at Google, would have had eight members and met four times over the course of 2019 to consider concerns about Google’s AI program.
Those concerns include how AI can enable authoritarian states, how AI algorithms produce disparate outcomes, whether to work on military applications of AI, and more.
Of the eight people listed in Google’s initial announcement, one (privacy researcher Alessandro Acquisti) announced on Twitter that he won’t serve, and two others are the subject of petitions calling for their removal — Kay Coles James, President of the conservative Heritage Foundation think tank, and Dyan Gibbens, CEO of drone company Trumbull Unmanned.
Thousands of Google employees have signed the petition calling for James’s removal.
James and Gibbens are two of the three women on the board.
The third, Joanna Bryson, was asked if she was comfortable serving on a board with James, and answered, “Believe it or not, I know worse about one of the other people.”
Altogether, it’s not the most promising start for the board.
The whole situation is embarrassing to Google, but it also illustrates something deeper: AI ethics boards like Google’s, which are in vogue in Silicon Valley, largely appear not to be equipped to solve, or even make progress on, hard questions about ethical AI progress.
Nearly half the board has resigned or is under fire
Google announced their AI ethics board at the end of March.
From the start, the board attracted criticism.
Many people were outraged about the inclusion of Kay Coles James, the Heritage Foundation President.
“In selecting James, Google is making clear that its version of ‘ethics’ values proximity to power over the wellbeing of trans people, other LGBTQ people, and immigrants,” argues an open letter signed by more than 1,800 Google employees.
A particular cause for concern was James’s stance that the trans rights movement is seeking to “change the definition of women to include men” in order to “erase” women’s rights.
“Google cannot claim to support trans people and its trans employees — a population that faces real and material threats — and simultaneously appoint someone committed to trans erasure to a key AI advisory position,” concludes the open letter.
Others called on Google to remove Dyan Gibbens from the board.
Gibbens is the CEO of Trumbull Unmanned, a drone technology company, and she previously worked on drones for the US military.
A year ago, Google employees were outraged when it was revealed that the company had been working with the US military on drone technology as part of so-called Project Maven.
With employees resigning in protest, Google promised not to renew Maven.
On 30 March, Alessandro Acquisti, the privacy researcher, announced his resignation from the panel, saying, “I’d like to share that I’ve declined the invitation to the ATEAC [Advanced Technology External Advisory Council].”
“While I’m devoted to research grappling with key ethical issues of fairness, rights & inclusion in AI, I don’t believe this is the right forum for me to engage in this important work.”
Even before the outrage, this panel was not set up for success
But the collapse of Google’s panel and the controversy over its make-up almost obscures a deeper problem: This was not an entity set up to do a good job.
Google’s announcement states that the panel would serve over the course of 2019 and meet four times.
That’s just not very much time together, given the complexity of the issues members will be advising on.
It’s not enough time to hear about even a fraction of Google’s ongoing projects, which suggests the board won’t be giving advice on those.
Second, the board positions are unpaid.
Some have contended that a paid oversight committee would be worse, because board members would be indebted to Google, but others think unpaid board positions advantage the independently wealthy.
These critics see the unpaid positions as another sign that Google isn’t taking the AI ethics board very seriously.
Next, the ethics panel — as has been the case with ethics panels at other top tech companies — does not have the power to do anything.
It’s very unclear who, if anyone, at Google will rely on these recommendations and which decisions the board will get to make recommendations about.
Overall, it’s not clear whether the panel will be used for guidance on internal Google matters at all.
What it definitely will be used for is PR.
If Google’s goal with the panel is to “be convincing to society broadly” without necessarily changing anything the company does, that’s not really AI ethics — it’s AI marketing.
And fundamentally, that’s what’s wrong with AI ethics panels.
Google is not the only tech company to have one, and while Microsoft’s AI ethics committee and Facebook’s centre for ethics research have not been embroiled in quite as much drama, they don’t have official decision-making power, either.
Ethical deployment of powerful emerging technologies involves tough decisions.
Will a company work with the Chinese Government on technology that aids it in its ongoing, horrifying campaign to imprison a million Uighurs?
If a facial recognition tool works better on white than black people, what does it mean to fairly deploy it?
If AI is creating and exacerbating inequalities, what’s the plan to tackle them?
All of those calls have to be made at the highest level of the company.
A better panel would contain both decision makers at Google and outside voices; would issue formal, specific, detailed recommendations; and would announce publicly whether Google followed them.
Neither Google nor anyone else appears actually comfortable with meaningful external oversight.
Neither Google nor anyone else seems to have a principled or systematic way to handle the power it has stumbled into.
The brouhaha has convinced me that Google needs an AI ethics board quite badly — but not the kind it seems to want to try to build.
* Kelsey Piper is a staff writer at Vox’s Future Perfect. She tweets at @KelseyTuoc.
This article first appeared at www.vox.com.
