Join Bridge Winners
All comments by Jean-Baptiste Fantun
1 2 3
You are ignoring the author of this comment. Click to temporarily show the comment.
For obvious reasons I will answer your questions in private Richard :)
July 9
You are ignoring the author of this comment. Click to temporarily show the comment.
Yes Richard this is definitely our approach.
July 9
You are ignoring the author of this comment. Click to temporarily show the comment.
Good move Alan, we would obviously be so happy to contribute ! (Thanks R Willey and M Bodell for mentioning NukkAI by the way)


NukkAI is a private lab of Artificial Intelligence, based in Paris (but proactively engaging in an international approach), whose scientific objective is to build a program that could outperform champions at Bridge and explain its decisions. The innovative methods we develop are likely to have an impact on different fields: healthcare, agronomy, insurance, etc

The NukkAI project has two « legs »:

- the « left » leg is dedicated to research (focused on bridge) and has taken a huge step in just one year. Starting in May 2018, we have already gathered a strong team of prominent AI researchers, young engineers and bridge experts (B Moss and P Cronier for example) and are moving fast.

- the « right » leg is dedicated to applications but taking a step with this leg required a credible scientific team which we now have gathered thanks to the bridge challenge. So now we are ready to tackle different verticals, developing POCs with different partners; 2 first applicative fields are humanitarian (AI for good) and industry.

On the research/bridge part:

- our innovative algorithms mixing Deep-Learning, Reinforcement Learning and other paradigms are already beating state of the art bridge bots and on some restrictions of bridge we are approaching human level

- we have already realized POCs where we « open the black box », ie our bots take an optimal decision and we can plug a module that « explains » the rationale for the decisions

- we have also designed tools for research purposes that might some day prove useful in a bridge context


Please feel free to make contact (jbfantun@nukk.ai) for more information. The website (www.nukk.ai) will be updated this summer. We try to answer questions and provide inputs (presentations, links to articles, videos, etc) to the people who get in touch. We are really extremely busy now completing the current fundraising (we are raising 1M€ to accelerate on research and develop applications) and still have not thought about the way to federate all the people who offered to contribute one way or another to this beautiful challenge; but we are currently thinking this over and will hopefully come up with a proposition to our numerous contacts all over the world.

Meanwhile, we will be in Las Vegas from July 20 to July 30, to play in the NABC, have discussions with bridge champions (giving them updates about our bridge bots) and potential investors (the project is more and more international and attracting investors in the US). And we are available to grab coffee/beer and discuss the NukkAI project with anybody from BW:)
July 9
You are ignoring the author of this comment. Click to temporarily show the comment.
Bridge is dying and it is a game for old people…in Western countries. Looking at Under 21 World Championships might convince us that the future of Bridge lies in China, Singapore, etc
May 5
You are ignoring the author of this comment. Click to temporarily show the comment.
At NukkAI (www.nukk.ai) we are developing (for research purposes) bridge bots that can in particular adjust their settings to their human partner. We are thus working on such psychological profiles (adapting them to bridge) and will soon launch a study to define categories, determine which categories are overrepresented among bridge champions, how the profile influences the decision making process, etc

If you are interested in participating in the study or just in having more info, please email me via BW or jbfantun@nukk.ai

Thanks Eric for bringing up the topic !
May 2
You are ignoring the author of this comment. Click to temporarily show the comment.
Well at NukkAI (www.nukk.ai) we do build bridge-playing bots relying on modern AI methods but I can assure you they are no threat to anybody, except to their partner at the moment :)
April 24
You are ignoring the author of this comment. Click to temporarily show the comment.
Yes exactly but we do not admit it publicly :)
Jan. 25
You are ignoring the author of this comment. Click to temporarily show the comment.
Alphago's victory against Lee Sedol has not made Go irrelevant.

Have you ever thought about the reason why researchers tackle games ? Because there is a limited number of rules, because data are clean, because the algos you develop to solve a game prove useful in real-life applications…
Jan. 25
You are ignoring the author of this comment. Click to temporarily show the comment.
This looks like a new great achievement by DeepMind.

As for Bridge, I do not know about DeepMind but at NukkAI (www.nukk.ai) we are currently focussing on the Bridge challenge and moving fast thanks to our international team mixing AI researchers, engineers and champions like Brad Moss or Philippe Cronier.

Of course the difficulties at Bridge lie in the incomplete information but also in the collaboration aspect. What is more, we want to build a program that can “explain” the rationale for its decisions because this element is paramount in the context of future applications.
Jan. 25
You are ignoring the author of this comment. Click to temporarily show the comment.
There are lots of different issues involved here:


- BBO-Funbridge: I have no clue about the details of the merger but in a context of rising competition I like the idea that American and European companies join their effort to build a leading actor on the computer bridge scene.

- I know even less about the Ginsberg-Al Levy issue but what I know is that Al is very dedicated to the development of computer programs and more broadly research projects related to bridge.

- I cannot agree more with Jurgen’s assertion that the computer bridge is an extremely interesting research topic and that scientific advances in this field may have a huge impact much beyond bridge. This is what the NukkAI project (www.nukk.ai) is all about (thanks Pierre Schmitt and Alvin Levy for mentioning Veronique Ventos who cofounded this project with me). We are indeed using the game of bridge as a « sandbox » to develop innovative AI techniques because the methods that allowed to crack the game of Go are likely to be much less efficient in a context of incomplete information. And In real-life (healthcare for example) the information is almost always incomplete so these methods designed to deal with the bridge challenge will certainly find applications in many fields.

Some inputs about NukkAI which is definitely international (bridge champions like Brad Moss or Philippe Cronier have already joined ship and we have had very interesting discussions in Orlando with Kit Woolsey or Eric Rodwell for instance).

- Inductive Logic Programming conference in Ferrara, Italy (August 2018): we have presented the first NukkAi paper (https://link.springer.com/chapter/10.1007/978-3-319-99960-9_5) at an international ILP conference gathering researchers from this symbolic AI community; it was a great opportunity to show that Bridge is a killer application for ILP and a good starting point for academic collaborations.

- World Bridge Championships in Orlando, USA (September 2018): our presentation was very well received by the Executive Committee of the World Bridge Federation and the Daily Bulletin displayed a great article about NukkAI.

- FranceisAI meeting in Paris, France (October 2018): Veronique was invited to give a keynote (https://www.youtube.com/watch?v=V97fFAbahn0) that may give a good idea of the project.
Dec. 28, 2018
You are ignoring the author of this comment. Click to temporarily show the comment.
We have defined some milestones but no date has been precisely settled yet.
Sept. 18, 2018
You are ignoring the author of this comment. Click to temporarily show the comment.
Thanks Richard ! We were recently having a heated argument about simulations and I said “if only R Willey was here… !! ”. Will sure let you know if we come to Boston or NYC.
Sept. 18, 2018
You are ignoring the author of this comment. Click to temporarily show the comment.
Of course it is all the whole case is about.

Just a little story: weeks after the FN case was exposed, there was a French Bridge competition (Finale Nationale d'Interclubs) whose participants included the Z team and some prominent local teams. One would expect that Mr Z would step down from this competition because his team had qualified for this Finale Nationale playing with FN. One would at least expect something like an apology for sponsoring a team that have deprived so many players (especially in France where they played a lot) of bridge titles. Nothing like this happened and many players were shocked by both the presence and the attitude of the Z team: I suggested to some fellow players that all teams should refuse to play against the Z team in protest of the situation but very few would agree with me (I did not have a chance to do that because my team did so badly in this competition that we never faced the Z team), most arguing that “Mr Z was a victim of FN” and that “what was happening was terrible to him”.

What this story tells us:

- Mr Z poses as a victim who has been betrayed by FN and many think he should be pitied, especially in the light of “all his efforts to promote our game”

- Under any circumstances will the Z team disavow the championships and the CAS decision will help justify this position


Finally, the Z team and the Monaco Federation have not been openly supportive of FN but just after the CAS decision M. Allavena, the President of the Monaco Federation, commented on Facebook: “FN found not guilty by the CAS, what great news !”…
Jan. 17, 2018
You are ignoring the author of this comment. Click to temporarily show the comment.
Just shameful and very bad for bridge image.

I am told that a Facebook post showed the Monaco federation rejoiced at such news…How long before FN join the Z team again ?
Jan. 11, 2018
You are ignoring the author of this comment. Click to temporarily show the comment.
Mike, please read the article and you will understand that when very few simulations are run (as is the case for Wbridge5 and its competitors), the choice of the seed does matter and it is proved empirically in the paper.

You may think that the paper is not scientific, then you know a lot more than the scientific committees of international AI Conferences like ICTAI. Good for you.

The beauty of this controversy is that the whole team involved in this paper decided finally that the usual approach (Monte Carlo simulations coupled with double dummy analysis) was not going to allow the machine to surpass humans and that they needed to design something very different. So there is more to come and more fun on BW !
Dec. 28, 2017
You are ignoring the author of this comment. Click to temporarily show the comment.
Thanks Richard, I will forward the comment to V Ventos :)
Dec. 27, 2017
You are ignoring the author of this comment. Click to temporarily show the comment.
Ok Richard, my bad, I thought someone who knew well Bridge PRNGs would also know for sure that:

- Bridge PRNGs ARE borked (but I would be happy to see one that is not)
- Although WBridge5 is the one which runs the biggest set of simulations, it is still much too short (approx 120)

What V Ventos and O Teytaud have done was boost an existing AI (WBridge5) with all its flaws and their conclusion was that setting up an AI able to outperform humans needed designing a whole new architecture involving new AI techniques.
Dec. 27, 2017
You are ignoring the author of this comment. Click to temporarily show the comment.
The authors are working on a global project involving Ai and Bridge and you can certainly email V Ventos to have details about the project (including other papers).
Dec. 27, 2017
You are ignoring the author of this comment. Click to temporarily show the comment.
I envy those who have time to spare criticizing the grammatical errors in a paper written by a non English-speaking researcher, instead of trying to understand the scientific message.
Dec. 27, 2017
You are ignoring the author of this comment. Click to temporarily show the comment.
Yves Costel, one of the co-authors of the paper, is indeed the developer of WBridge5 that had not won the World Championships for 8 years until its boosted version won the Championships in Wroclaw (2016) and in Lyons (2017).
Dec. 27, 2017
1 2 3
.

Bottom Home Top