2017 Watson AI XPrize Top 10 Revealed at NIPS

　　The $5 million IBM AI Watson XPrize is revealing the top 10 finalists for the 2017 round and awarding a total of $15,000 in prize money to the top two finishers today (Dec. 8) at the Neural Information Processing Systems conference (NIPS 2017; Long Beach, Calif.). Amiko AI (Milan), which is upgrading respiratory care with sensor technologies and digital health tools, has come in first and is being awarded the $10,000 top prize for this year. The $5,000 second-place prize goes to aifred Health (Montreal), which is using deep-learning algorithms to personalize treatments for depression. The two top finishers also were scheduled to present detailed descriptions of their projects on stage at the event.

　　“The 32 judges — who are all independent of IBM, as am I — narrowed the 147 first-year contestants down to 59 second-year contestants [based on] who has made the most progress on the most helpful-to-world-society projects,” Amir Banifatemi, prize lead for the IBM Watson AI XPrize and managing partner of K5 Ventures, told EE Times in advance of the announcement at NIPS.

　　Here are the other top 10 finishers, in alphabetical order:

• BehAIvior (Pittsburgh) is combining data from wearables and smartphones to create an early-warning system that predicts addiction relapses — especially overdoses — with the intent of preventing them.

• Brown University's Human Centered Robotics Initiative (HRCI; Providence, R.I.) is creating a three-phase program to identify the social and moral norms that robots should be designed to internalize.

• DataKind (New York) is developing artificial-intelligence models using high-resolution satellite imagery that can help alleviate poverty in underdeveloped regions by monitoring crops for disease while they can still be saved.

• Deep Drug (Baton Rouge, La.) is working on AI drug design software that learns from both the successes and the failures of previous clinical trials to shorten the development time for new, more-targeted drugs.

• The EmPrize team at the Georgia Institute of Technology’s Design & Intelligence Lab (Atlanta) is aiming for smart virtual tutors that will answer questions, provide feedback, and perform other functions for online education.

• EruditeAI (Montreal) is creating a free peer-to-peer math tutoring platform to match students who are struggling to understand a given mathematical concept with students who have demonstrated proficiency in that concept.

• Iris.ai (Oslo, Norway) is automating a systematic mapping solution for scientific papers that will help AI researchers with the literature discovery phase of their projects.

• WikiNet (Quebec City) is working on a solution that learns from past environmental-cleanup efforts to provide expert recommendations for cleaning up other contaminated sites.

　　The 59 teams selected to move forward from 2017 to 2018 come from Australia, Barbados, Canada, China, France, Germany, India, Israel, Italy, Norway, Poland, the United Kingdom, the United States, and Vietnam. The categories their projects cover include Health & Wellness, Learning & Human Potential, Civil Society, Space & New Frontiers, Shelter & Infrastructure, and Energy & Resources. The criteria used to assess the projects include the standards they intend to set; the performance and scalability of their application; and, most important, their potential to achieve an exponential societal improvement. “The judges this year recognized those teams that have emphasized man-machine collaboration and are furthest along in their projects,” said Banifatemi.

　　In a nod to the rapid pace of breakthroughs in AI, the AI XPrize added a wild-card round this past fall to accommodate teams working on concepts that were not foreseen in the competition’s first year. The top 10 finishers among the wild-card teams that applied for inclusion this year will inflate the total field of contenders to 69 in 2018. A second wild-card round will be held next year. In an interview with EE Times when the first wild-card round opened, Banifatemi said that, based on how many wild cards are approved to compete this year and next, “we expect to have half of the total teams in competition by September 2018 moving into 2019.” No wild cards will be added in 2019, and at the end of that year the field will be halved again.

　　Further prize money will be awarded to the top 10 finishers in 2018 and 2019, with the three top-10 rounds (2017-2019) collectively accounting for $500,000 of the $5 million total allotted for the XPrize. In 2020, at the Grand Prize event on the TED2020 stage, the remaining $4.5 million will be awarded to the top three finalists: $3 million to the first-place finisher, $1 million for second place, and $500,000 for third place. The third-prize winner will be selected with the help of voters at TED2020.