IBM Watson AI XPrize Adds Wild-Card Round

　　The $5 million IBM Watson AI XPrizecompetition, which kicked off last year and will end in 2020, was the first of the XPrize contests (14 since 1995) to have a contestant-defined “open” goal rather than a predetermined objective. Now it is also the first XPrize to add a wild card, giving new contestants until Dec. 1 to join the 147 teams that made the first-year cut.

　　“The total number of teams officially registered stands at 147, out of the total of 870 team submissions that were recorded from more than 9,000 initial interested requests,” Amir Banifatemi, prize lead for the IBM Watson AI XPrize, told EE Times. “Given the rapid pace of artificial-intelligence breakthroughs and the possibilities that AI opens to solve grand challenges, we wanted to ensure that teams with new ideas still had the opportunity to participate.”

　　The original teams still in the competition hail from 22 countries in total: Australia, Barbados, Canada, China, the Czech Republic, Ecuador, France, Germany, Hungary, India, Israel, Italy, Japan, the Netherlands, Norway, Poland, Romania, Spain, Switzerland, the United Kingdom, the United States, and Vietnam. Their projects are being evaluated not only for their efficacy in addressing AI challenges but also for their potential social, ethical, and technological impact. Imbuing AI with the ability to understand human emotional cues, for example, could have implications beyond the AI’s cognitive-computing capabilities.

　　The IBM Watson AI XPrize is also the first XPrize to leave the goal open to the contestants’ discretion, and as a result the teams have proposed ideas for solving problems across a wide swath of disciplines. “Energy Efficiency” projects would reduce greenhouses gases and makes landfills smart at separating recyclables. “Health and Wellness” investigations look to head off mental health problems, diagnose an infant’s crying, and improve sleep. “Learning and Human Potential” projects aim to reinvent computer coding, personalized learning, peer-to-peer tutoring, and scalable learning to achieve universal worldwide literacy. Proposals for “Improving Society” would automatically flag “fake news” in social media and get legal information to victims at little cost. “Shelter and Infrastructure” projects aim to meld social development with satellite imagery, predict disasters, manage traffic flows in cities, and assess the structural health of buildings. “Space and New Frontiers” explorations seek to develop neurologically inspired models and automatically propose hypotheses.

　　“We have been impressed with the level of variety and domain focus so far. Teams are very diverse, [hailing from] startups, academia, large corporations, nonprofits, and more,” Banifatemi said. Among the “impressive” entries, he said, are AI applications to “detect crop disease in Ethiopia, detect illegal mining in Congo, model malaria-prone regions in India, predict psychiatric medicine effectiveness, automate project management at scale, [advance] triage emergency medicine, and monitor the structural health of buildings.”

　　The addition of the wild-card teams aims to widen the application domains even further, but the expanded pool will still be subject to the same annual culling process. “Each year, up to 50% of teams will move to the next round provided they reach their milestones and are selected by judges to move forward,” Banifatemi said. “Based on how many wild cards are approved to compete, we expect to have half of the total teams in competition by September 2018 moving into 2019.”

　　The judges also will be picking out favorites over the course of the multiyear competition, distributing $500,000 in total to teams for meeting their stated milestones with outstanding performance. The milestone awards will be made at the judges’ discretion rather than follow a strict policy. The finalists in 2020 will receive $3 million for first place, $1 million for second place, and $500,000 for third place at the Grand Prize event on the TED2020 stage. The conference attendees, including the online audience, will have a say in determining the final placement of the three winners.

　　Banifatemi described the awards system in detail: “We will have 10 teams eligible for the milestone prizes each year, and the top three will receive cash prizes based on the judges’ assessment. The milestone prizes in 2017, 2018, and 2019 are part of the prize purse that IBM has committed to. By the final round, in 2020, three teams will have been selected from the 10 milestone prize candidates in 2019. The judges will have already approved the top three, and the public will weigh in on the final scoring. The judges score’ and the public score will be taken into account for selecting the first-, second-, and third-place winners."

　　The next major event, in December, will be the announcement of which of the currently registered teams (roughly half of the current total) will move on to the next phase. In January, the judges will announce which of the wild-card teams will be added to the roster. In January 2019, the field will be halved again, with survival dependent on the standards proposed and on the AI’s performance, scalability, and — most of all ??— likely worldwide impact.