Earlier this week, I traveled to Paris to attend the Global Forum on Artificial Intelligence for Humanity (GFIAH). The by-invitation event featured one day of workshops addressing issues such as AI and culture, followed by a two days of panels on developing trustworthy AI, data governance, the future of work, delegating decisions to machines, bias and AI, and future challenges. The event was a part of the French government’s effort to take the lead on developing a new AI regulatory framework that it describes as a “third way”, distinct from the approach to AI in China and the United States. The French initiative, named the Global Partnership on AI, is particularly notable from a Canadian perspective since Canada is an active participant in the initiative and will host the next global forum in 2020.
Reflecting the French government’s prioritization of the issue, French President Emmanuel Macron delivered a major address. With several government ministers in attendance, Macron outlined the French vision of a new global alliance on AI regulation, took numerous opportunities to criticize global technology companies (and researchers that work for them), and urged holding more GFIAHs.
I will have more to say about Macron’s initiative and what it means for Canada, but this post focuses on the Global Forum itself. Global discussions on AI policy are absolutely essential. However, as the title to this post suggests, this Global Forum was not global. Rather, the overwhelmingly European and North American crowd was strikingly unrepresentative. The most obvious example occurred after the panel on AI and bias, which featured several presentations on bias in data sets that are often used in the development of AI systems. These included Françoise Soulié-Fogelman, who cited a well-known study that facial recognition programs work well with white males (1% error rate), but not for minority women (35% error rate); Kate Crawford, who emphasized the diversity crisis in AI, noting that women are poorly represented in major labs at companies such as Google and Facebook; and Laurence Devillers, who discussed how AI systems can reinforce stereotypes with 80% of coders being male and 80% of chatbots female.
Yet after a compelling panel on bias issues in AI, the very next panel was an all-male panel: 7 men, no women. Several people tweeted about the issue while the panel was underway. When the issue was raised as the very first question, the panelists did not directly address the issue, alluding to a missing female panelist who was unable to attend. Simply put, the only thing the all-male panel succeeded in demonstrating was that the diversity crisis is very real since it even manifested itself in an event purporting to focus on AI policy for humanity (it is worth noting that the best presentation at the two-day event – by far – was from Marzyeh Ghassemi, who joined the University of Toronto in 2018).
But the problems with the Global Forum was not limited gender diversity on panels. After I tweeted about concern about the manel, I received a text from a friend in South Africa asking if there had been any examples or discussion involving Southern Africa. I quickly responded that there was practically no representation from the global south, with no speakers currently based in Africa, Latin America, or Asia (beyond Japanese representation). In short, this was a global forum that neglected to invite the globe. If the GFIAH is to be relevant, it must be far more representative.
In fact, not only was the global south excluded, but so too were the major technology companies. There were a couple of company representatives on panels (representing Thales, a French company, and Siemens, a German company) and two speakers who have affiliations with labs run by Facebook and Microsoft, but the broader corporate perspective was almost entirely missing. Some may argue that that is how it should be, leaving it to academics and politicians to develop AI policy. Yet global policy requires inclusive discussions that bring together all stakeholders, including companies and the under-represented NGO community.
As AI policy becomes increasingly important, there is unquestionably a need for a forum that brings together global perspectives in the hope of developing ethical rules and standards that prioritize people. Without an effort to be far more inclusive, the inaugural, unrepresentative GFIAH is not that forum.