Building Next Generation AI : 𝝂Bridge
The game of bridge is the next milestone for AI :
We are working hard on this fascinating challenge
Why Bridge ?
Being a good bridge player requires :
- Depth of analysis
- Processing multi-source information
- Reasoning in incomplete information context
- Reasoning with constraints i.e. time limits
- Communicating with partner while concealing things from opponents
- Evaluating opponent's level and psychology
To build a bridge robot, we need to model very different forms of intelligence.
How it all started
Veronique, a senior researcher at Paris Saclay, started working on bridge in 2015 and published
several papers on the subject. Her methods helped bridge program
WBridge5 win twice the World Computer Bridge Championships in 2016 and
2017. She and some fellow researchers gradually convinced themselves
that bridge was no less than the next milestone for Artificial
Intelligence. She gathered scientists from different AI communities and
bridge experts to launch the international Alphabridge ( now 𝝂Bridge ) project that grew
In 2018, prominent entrepreneurs joined ship and NukkAI was born, a private AI Lab whose primary objective is to focus on the bridge challenge.
What we are building
To crack a game, a program has to play optimally ; but to solve it,
the program's strategy must also be explainable in human understandable
Our goals are twofold :
- create an AI that could win machine vs human challenges.
- build a system that will extract human-readable rules expliciting the reasons for decision-making.
In short, it is about solving the bridge game.
Is there something beyond bridge ?
Designing an AI that could outperform humans at bridge while
explaining its actions is very difficult. A success in this challenge
would have a huge impact on "real-life" problems and could definitely
be the next step towards a general AI.
But let us keep our feet on the ground and let us focus on our challenge.
Stay tuned because things are moving fast !