A.I. and parallel computing can develop business strategies and tactics better than the best human brains.
FEBRUARY 15, 2017, TORONTO ON – Daisy Intelligence Corporation’s inaugural hackathon, held on January 28-29, 2017 at the University of Toronto’s Faculty of Applied Science & Engineering, showcased the power of artificial intelligence (A.I.) and GPU-accelerated computing. The hackathon also exposed engineering and computer science students to the exciting field of modern A.I. The event was sponsored by NVIDIA, a world-leading GPU hardware manufacturer, and Cogeco Peer 1, a tier-1 data centre company.
For 24 hours, students gathered in the Bahen Centre for Information Technology to develop the ultimate tic-tac-toe playing bot, executed on GPU hardware with five seconds of analysis to decide each move. The students were given the opportunity to pre-train their bots prior to the head-to-head double elimination tournament.
“We decided to task the students to develop a bot for ultimate tic-tac-toe because this represents a miniature version of the A.I. problem that Daisy Intelligence solves for its corporate retail clients: in a defined, time-constrained environment, make weekly/daily decisions that the clients will execute to maximize profitability,” said Gary Saarenvirta, CEO of Daisy Intelligence. “In the context of the hackathon, this means creating a program that within five seconds, decides moves it will make that will lead to a win. Similar to real-world retail, ultimate tic-tac-toe environments have practically infinite decision spaces – therefore it is not possible for a human mind or even a traditional computer approach to find the optimal strategy. A.I. combined with parallel computing can search more of the decision space than the human mind and would be expected to win every time.”
A total of 34 student teams registered for the hackathon. Almost all teams created a rules-based deterministic strategy executing in serial. The head-to-head tournament began with early rounds as best of three games with penalties for invalid moves or moves that exceeded the five-second decision limit. If a team made too many invalid or time exceeding moves in a single game, that game would be given to the opponent. Later round games were a best of eight with a draw being broken by number of squares won.
“The hands-on challenges offered by the Daisy Intelligence Hackathon stimulated creativity and rapid problem solving which are essential skills for our students to develop,” said Professor Deepa Kundur, Chair of the Engineering Science program at the University of Toronto. “Our programs promote technical excellence broadly and opportunities such as these provide fertile ground for students to learn to innovate under pressure.”
The winning team, Team 0.004184, was the one team who developed a deterministic strategy combined with a tree search. This team did not lose a single match, let alone a game, in winning the tournament. The other teams used deterministic logic with different heuristics. An analysis of the strategies from the submissions prove that a search strategy does better than a deterministic rules-based approach in all instances.
Daisy’s engineering team developed a bot to test the hackathon environment and played the tournament winner in a best of eight match with one second to execute each move. Daisy’s parallel random tree search algorithm was allowed to search a maximum of 500 milliseconds. Longer search times were also tested, and Daisy won all those matches 8-0 versus the tournament winner.
“We were surprised by what the students were able to accomplish in the allotted time. We didn’t expect every team to come up with a GPU-enabled parallel software strategy,” said Saarenvirta. “This hackathon is a great example of how even a rudimentary parallel A.I. method can defeat a human developed rules-based strategy every time. If we extrapolate this to the business world, A.I. and parallel computing can develop business strategies and tactics better than the best human brains. Not only do we expect financial benefits to the business space as A.I. grows its scope of influence, we also expect A.I. will eliminate much of the repetitive, quantitative and boring tasks that are part of today’s jobs.”
About Daisy Intelligence
Daisy Intelligence is an artificial intelligence software-as-a-service company that analyzes very large quantities of our clients’ transaction and operational data in order to make automated operational decision recommendations which our clients can immediately action to improve their business. Using our proprietary mathematical solutions and the Daisy A.I. based simulation platform, Daisy Intelligence analyzes 100% of the tradeoffs inherent in any complex business question and provides weekly, specific recommendations to help our corporate clients grow total sales, improve margins, reduce fraud and delight customers.
About University of Toronto Faculty of Applied Science & Engineering
The University of Toronto’s Faculty of Applied Science & Engineering is Canada’s premier engineering school and among best in the world. Our diverse community includes more than 5,400 undergraduates, 2,300 graduate students, 260 academic staff and nearly 50,000 alumni. Through innovations in engineering education, we prepare the next generation of global engineering leaders with strong technical foundations and professional competencies in multidisciplinary research, teamwork, leadership and entrepreneurship. Our faculty members are international leaders in research who collaborate across disciplines to address key global challenges, from new diagnostics and treatments for human diseases to wearable technology, smart cities and renewable energy.
For more information, contact:
CEO, Daisy Intelligence Corporation
905.642.2629 ext. 221
External Relations Officer, Division of Engineering Science
Faculty of Applied Science and Engineering, University of Toronto