Nukkai, the start-up pioneering a next-generation Artificial Intelligence, has announced a clear victory of its Artificial Intelligence pitted against eight World Bridge Champions over 800 consecutive deals, divided into 80 sets. With 83% of games won by the machine, out of a total of 80, the Nukkai robot has shown for the first time in the history of bridge that the AI can outperform the best human players during the essential stage of the game [1].
Co-founded by researcher Véronique Ventos and tech engineer Jean-Baptiste Fantun, Nukkai selected the game of bridge for the experiment as it mirrors many of the decision-making challenges we face in real-life.
Cédric Villani, French MP and 2010 Fields Medallist, comments, “Looking beyond this feat that will go down in history for bridge and AI, we are gaining fascinating perspectives on the understanding of probabilistic reasoning using hidden information, and on the explainability of algorithms. This is a superb French success story!”
The Challenge
On 24th and 25th March, Nukkai’s AI played against eight undisputed World Bridge Champions [2] in Paris. They represented six nationalities and hold 16 World Champion titles in total. These champions competed against Nukkai’s AI in a tournament involving 100 deals for each of them. And the final score reflects the AI’s excellent performance even compared to the most seasoned players.
Unlike chess or Go, bridge is more resistant to AI because it is a multi-player game with incomplete information and whose rules impose some explainability. Nukkai is the first company in the world to have developed a robot that is able to reach a superhuman level during the essential phase of the game, thanks to a next-generation AI.
According to eminent Pr. Stephen Muggleton (Imperial College) who gave the closing event speech, “Many of the researchers to whom I sent the (live event) link were full of praise for your amazing achievement.”
Next-generation AI
Nukkai’s AI is an explainable, collaborative and energy-efficient AI. It thus differs from AI and the traditional “black box”, which raises questions of transparency, human/machine interaction and energy consumption.
Jean-Baptiste Fantun, says, “One of the greatest challenges of AI is how to foster trust in systems powered by AI. Humans and businesses are increasingly uncomfortable with “black box” type AI algorithms that do not indicate how decisions were made.”
Explainability: Nukkai has developed a tool that analyses a game and “explains” the strategy employed. It can then transfer the robot’s skills to the human or show human how their strategy was inferior to that of the robot.
Collaboration: While the “black box” type AI complicates any interaction with the human, Nukkai’s AI can interact with it; having the ability to explain its choices and decisions. And it is the human who makes the decision, following suggestions made by the machine.
Energy sobriety: Developed with the support of the CNRS (Le Centre National de la Recherche Scientifique) – which provided access to the most powerful supercomputer in France, the Jean Zay – Nukkai’s AI is 200,000 times less energy-intensive than the solution used to beat the Go World Champion.
Reshuffling the cards of AI
Véronique Ventos, named by Forbes the third most influential woman in the world in the field of AI, adds, “AI is usually based on digital methods linked to neural networks, and to deep learning, following the progress made in computer power. Unfortunately, these approaches do not provide explanations that a human can understand. Nukkai’s AI relies on innovative hybrid methods: having proven itself with a game as difficult as bridge, there should be no difficulty finding applications in areas with strong economic potential.”
Nukkai’s AI (applied to bridge where the opponent’s cards aren’t seen [3]) is prime for development in many verticals because we never have all the information at our disposal in real life. However, Nukkai has already started applying its methods to other areas:
- Education: Nukkai has developed a product called Kartoon, offering primary school pupils adaptive exercises to enhance their cognitive and mathematic skills while providing rich feedback to the teacher. It was rolled out in September 2021 across several schools in France.
- Aeronautics: Nukkai has developed an initial tool that is particularly useful for airlines and will be signing a major agreement with a major player in this field soon.
- Cybersecurity: The methods developed by Nukkai are particularly effective in this sector, where experts are inundated with alerts and need an interactive tool that allows them to distil information and automate as many processes as possible.
- Defence: Nukkai has been working closely with a major French player in the defence sector for several years.
Nukkai shares its success with the entire French Tech ecosystem. The start-up benefitted from financial support from the BPI (French public investment bank), including deep tech development aid of €1 million, the CNRS (which provided the Jean Zay supercomputer) and several major academic partners [4].
Following President Emmanuel Macron’s recent commitment to AI, such a global event illustrates the first concrete results of France’s proactive AI policy, respecting the values of trust and the ambition to deliver “AI for good”.
About Nukkai
Co-founded in 2018 by researcher Véronique Ventos and technical engineer Jean-Baptiste Fantun, Nukkai is a French start-up that is pioneering a next-generation Artificial Intelligence. Why is it next-generation? Because Nukkai’s AI is the only one that is explainable, collaborative and efficient. Using bridge as a sandbox, Nukai’s AI aims to support sectors such as education, industry, aviation and cybersecurity in their economic and digital challenges. Nukkai’s next-generation AI constantly interacts with humans, it has the ability to explain their choices and decisions, and it is also energy efficient, a first in the world of Artificial Intelligence.
[1] That is to say, the playing of the cards corresponds to the phase after the bidding phase, when the team playing the contract must win at least the number of tricks it has committed to make.
[2] The World Champions include: Bénédicte Cronier (France), Thomas Bessis (France), Brad Moss (USA), Nevana Senior (United Kingdom), Sabine Auken (Germany), Roy Welland (USA), Anna Gulievich (Russia), Mikael Rimstdedt (Sweden).
[3] Bridge is a game that uses incomplete information (we only see our cards and not our opponent’s) and the methods used for Go are ineffective for this type of game.
[4] Véronique Ventos, co-founder of Nukkai, has spent her entire career at Paris Saclay, Professor Tristan Cazenave, who invented robot AI, holds an AI chair at Paris Dauphine, and Nukkai collaborates with the 3IA institutes in Grenoble. and Toulouse, the cognitive science laboratory of ENS Paris, Paris 13 (Villetaneuse), Caen, etc.
