G. Role of incentives and prizes
(RFI question 10)

Last updated July 28, 2016

As the fundamental building blocks of AI improve, so too should the incentives that inspire next-generation, people-centered systems design. As an example, IBM established a $5 million AI XPrize for the best use of AI system to empower teams of people to tackle the world’s grand challenges. IBM is developing additional scientific challenges for the AI research community.

Multi-year technology competitions driven by prizes and other incentives have proven an effective means to focus innovation efforts as well as shape markets for rapid adoption of new technologies. AI promises both revolutionary advances as well as broad application to many of society’s most intractable challenges such as education, healthcare, unemployment, and sustained economic growth. The government should sponsor more competitions targeted specifically at these challenges.

Technology Competitions: Government-sponsored prize-driven competitions have a long and successful history in motivating innovation, especially in conquering technical barriers that inhibit large-scale economic and societal transformations. The first such competition, “An Act for Providing a Publick Reward for Such Person or Persons as Shall Discover the Longitude at Sea”, was chartered by the British Parliament in 1714 to accelerate the discovery of navigation technology, namely a reliable means to measure longitude, that would go on to revolutionize marine commerce and exploration. More recently the DARPA Challenges in 2004, 2005 and 2007 focused national innovation on the development of autonomous vehicles, which has culminated in a number of semi-autonomous and fully autonomous driving features being incorporated into a range of commercially available vehicles. The DARPA Robotics Challenge, starting in 2012, is currently driving innovation in robotics technology toward practical applications in emergency and rescue operations in environments too dangerous for humans. In the commercial sector, organizations such as XPrize have successfully used prize-based competitions to drive innovation in a number of disparate fields, from commercial space travel to global literacy to breakthroughs in clean water availability. The demonstration of Burt Rutan’s SpaceShipOne as the first commercially reusable spacecraft culminated over 8 years of intense competition in the Ansari XPrize. This process has now spawned multiple large-sale commercial space transportation businesses with economically successful and technically impressive innovations in launch, control, re-entry, recovery and reuse. This year, IBM and XPrize have announced the IBM Watson AI XPrize with the intention of motivating breakthroughs in Cognitive Computing.

Criteria for Successful Challenges: As shown above, the multi-year, prize-driven competition format can be effective in focusing research and development activity across the country and the world. There are several keys to success in creating and running these challenges. Selecting a challenge that is beyond the state of the art but within 5-8 years of achieving with focused activity and innovation is paramount. Any challenge taking longer is unlikely to hold the attention of the research community or the market. We like the XPrize’s criterion of “audacious but achievable.” Incenting the competitors with prize money is important, but even more than the actual prize amount is the leverage gained in the market by the eventual winner, often 10 to 50 times the prize amount. This leverage can be delivered by publicity of winning the competition, clear market leadership as the winner, and market shaping incentives like preference in consideration for large government contracts, among other advantages. But the most critical success factor for any competition is quite straightforward: there must be competitors. Challenge selection must carefully consider the population of potential competitors, their motivations and staying power, as well the market timing for the winners if the challenge is achieved. This is also true of the sponsoring organization, that they will be able to sponsor and run the competition for the years it will take to complete. This implies that government sponsors be fully funded up front to operate these competitions across multiple election and administrative cycles.

The criteria above can serve for a challenge competition in any area of science. So what is different about AI competitions? First, there are a wide range of various AI competitions in play already, from various versions of the original Turing Test to General Video Game Playing and even multiple events inspired by the Olympic Decathlon. These competitions vary widely in their challenge domain, degree of participations, likelihood for near-term achievement, and level of incentives. If the government was to create one or more AI challenge competitions, it would need to compete for attention and participation with the many that already exist. A real opportunity for the government in this situation is to sponsor a competition that captures the imagination of both the public and the research community. Selecting the topic for the challenge becomes the primary consideration. We have already in this submission discussed the revolutions possible in childhood education, skills training, skill augmentation, manufacturing, healthcare, scientific discovery and other fields if certain barriers in AI are overcome. We believe that multiple challenge competitions should be sponsored within these opportunities, and IBM looks forward in both facilitating and participating in them.

Back to summary.