Next, Stefan Wolff talked about ethnic conflicts and civil wars. He wondered if all the pictures of current conflicts he showed on screen were to be pictures of the past. He argues there has been a decline in the number of conflicts since the 90s, due to institutional design, leadership and diplomacy. The subject was interesting but frankly the presentation a bit underwhelming because Stefan kept reading his notes too closely.

Then Eric Berlow did a short talk about the difference between complex and complicated. He showed how it was possible to find simple answers by raising complexity. He took as an example the NYT-leaked diagram for reaching popular support for the Iraqi government. An unholy mess of arrows but as he increased complexity and represented the graph in 3D, a simpler way to resolve dependencies emerged. Nice one!

William Perrin, the ex-CTO of the Blair administration from what I understood, talked about his neighborhood in London, which although very central seems to have been deserted by law enforcement. He showed how he managed to return to some kind of local civic democracy by launching a low-tech blog, and making people join forces. He argued governments should pledge to make the Internet the first means of communication with people.

Mallika Sarabhai, who opened TEDIndia, danced and sung. Some parts of it I liked, some others not.

Finally, Zainab Salbi talked about her childhood under missile strikes. She emphasized the importance of women in wars and most importantly in peace.

This last session closed TEDGlobal 2010, we unfortunately had to leave without going to the final BBQ party because of travel constraints. Next year maybe!

May be linked to his (albeit very thin) ties to the currently hugely unpopular Sarkozy, may be linked to his latest BP speech which was widely panned by the media. I’ve also recently read more solid criticism of his failure to live up to his campaign promises.

My humble take from abroad is that of course the hopes were too big when he was elected, but most people seem to forget now that they were already aware of it then. I don’t believe that just one president in less than half its first term could reshape completely the biggest political system of the world. But we’re undoubtedly on the good, or at least on a way better track than before. It’s just too big an effort and we’ll need more than one Obama to achieve it.

So IMHO, now is a bit too early to stop believing in the first one we get.

In the meantime, my side project TEDx PARIS is finally becoming big, tomorrow is the first public conference “à la TED” in Paris. We’ve got an amazing lineup of speakers for the 650 attendees, and that’s without the many surprises we’ve prepared

It has been soldout for one month but we have prepared a live video stream that will begin with the conference at 14:00 Paris time. It will be available on www.tedxparis.com/weblive/, liberation.fr and france24.com

Wish us good luck !

Click here to view the embedded video.

That’s it! For those interested in the source, it was moved on the official FON repository. Feedback welcomed!

The idea is to allow Foneros to listen to our Creative Commons-licensed radios with an USB sound card plugged in the new USB port of the Fonera. These small sound cards can be found for as little as $1 on eBay, we already ordered a couple of them for the office.

So the application is still in prototyping stage but we’re now familiar with the development environment of the Fonera and have a much better idea of how the app can work.

A few screenshots :

That’s it! We still have to figure out how to start the radio automatically when the Fonera starts, add an authentification layer and maybe some sort of volume control.

We have checked in the code on our googlecode repository, feel free to comment!

• We launched Jamendo.net blogs! This blog is now actually hosted on jamendo.net, which gives me the ability to test the platform myself
• I really like the iPhone wordpress client
• We’re going to the Web2 Summit 08 with Laurent, Pierre and Eliane next week! Timing couldn’t be better because we’ll also witness the U.S. Election Night on November 4th. Finally we won’t be the only ones around wearing Obama t-shirts. Anyway, it’ll be really interesting and I’ll try to blog what I can for my future children to read

Let’s begin! The problem is, we have a list of weighted tags for each album and we want to compute for each album the list of album that have the most similar tags list.

Here is a sample tag/weight list for the Manu Cornet album :

Let’s call this base album “A1″ and the similar albums we’re looking for “A2″. We can consider that “20″ is the maximum tag weight.

A quick list of requirements for the algorithm :

• Shouldn’t be symmetric
• Should penalize A2s if they lack one or more of the most important tags of A1
• Should be quite fast, even if we’re not going to run it on the fly (everything will be computed once a day)

Martin told me yesterday about the nice app “Grapher” which is bundled with OS X. I’m going to use it to write the formula.

My first idea was, for each tag of A1, to compute a number Z between -1 and 1 and add them all. Let (x,y) be the weight of the tag in A1 and A2 respectively. The most obvious algorithm would be :

z=20-ABS(x-y)

The axis on the left is Y, on the right it’s X and the elevation is of course Z.
The window is (0,20),(0,20),(-20,20)

Easy improvement : multiply by x/20

z=(20-ABS(x-y))*x/20

Quite good but there is a problem over the X axis. If y=0, z should always be < 0. Another easy solution : add min(0,y-x)

z=(20-ABS(x-y))*x/20+min(0,y-x)

New problem : x=20, y=10. In these conditions z shouldn’t be nul. Let’s multiply min(0,y-x) by (1-y/20).

z=(20-ABS(x-y))*x/20+min(0,y-x)*(1-y/20)

Better, but not good enough yet. (1-y/20)^4 ?

z=(20-ABS(x-y))*x/20+min(0,y-x)*(1-y/20)^4

Well, it looks quite good to me! Maybe we could tweak some curves a bit but let’s go for this one for now.

Here are the results for the Manu Cornet album :

I am not so happy with the results. Maybe there need to be another negative zone for tags that are in A2 but not in A1 after all.

Let’s reduce the window size to (0,1),(0,1),(-1,1) and take those “20″ factors out. Now much easier, here is what I’ve come up :

Looks good too but let’s compare the results :

Quite funny that both queries take the same time to process : the formula doesn’t seem o be the bottleneck. Results are a bit better but I think there will still be some tweaking needed in the future. Maybe reduce the negative zones.

Anyway, now I’m going to implement this and the new data should be online in a few minutes. Don’t forget to check our the “Similar albums” slider on your favourite albums pages!

This is a random radio of all the albums. Please report any issues with it as it will be official in a few days!

Update : Once again, thanks for irpyC from jamendo for his nice GUI !

&amp;amp;amp;amp;amp;amp;nbsp;

The presentation is focused on how and why jamendo and the artists use the Creative Commons licenses.

That’s it… See you soon Amsterdam and special thanks to all the Knowledgeland team for putting this event together, you guys rock!