don’t represent a grand challenge. In time, we expect
some of these teams will show empirical promise, but
without preliminary evidence they are unlikely to
advance.
Ethics and the Future of AI
The most challenging aspect of running an open-
ended competition for artificial intelligence is the
capacity for AI systems to solve global challenges (see
table 2 for team geographies), while also introducing
novel and unforeseen trade-offs. Teams competing in
the AIXP may deploy mental health dialogue agents,
medical recommender systems, and other technolo-
Table 2. Home Countries, Counts, and Advancement Rates for Competing Teams.
Country Team Count Advancing Count Advancing Percent
Barbados 1 1 100
Israel 1 1 100
Norway 1 1 100
Poland 1 1 100
Canada 20 11 55
UK 6 3 50
USA 71 30 42
China 6 2 33
Italy 6 2 33
Vietnam 3 1 33
France 7 2 29
Australia 8 2 25
Germany 4 1 25
India 5 1 20
Netherlands 2 0 0
1 0 0
Ecuador 1 0 0
Japan 1 0 0
Romania 1 0 0
Spain 1 0 0
Switzerland 1 0 0
Czech Republic
gies where the betterment of the many does not pre-
clude harm to a few. AIXP judges serve as arbiters of
global beneficence, but there is currently no expert
body that has a global process for recommending
procedures for deploying and monitoring AI systems.
While the IBM Watson AI XPRIZE has the resources
to review AIXP teams, a near future with ubiquitous
AI requires review methods that scale beyond formal
committees of the world’s leading experts. Many
organizations are working to fill the void of formal
process. Major corporations developing AI products
formed the Partnership on AI2 as a joint effort with
civil society organizations. Academics and engineers
drafted principles and standards for the ethical devel-
