Monday, September 17, Day 1

9-10.20am

First five presenting companies.

Session Topic: Search and Discovery

Powerset looks pretty interesting. Always tough to beat Google.

Searches;

- Who mocks Tony Blair

- Mike Arrington - using Powermouse.

The are using parsing to get better results. As well as letting user vote for the search results that they like. You get karma points.

Up now - Cognotive Code.

A pic from Powerset

Quick intro from Mike and Jason;

Cognitive Code had some language recognition and AI stuff.

Cast TV was video Search. Looked pretty good, but not earth shattering.

Good demo though.

Faroo up now.

Peer to peer search engine.

Cast TV Pics

Karoo was an interesting product but the guys accent was a bit strong to understand.

Viewdle also have video search.

Using Britney Spears.............. let's move on!

Pics from Faroo;

Viewdle pics.

Not bad demo. Using facial recognition to tag videos.

The companies are now on stage with the panel experts.

(I thought Marrissa Meyer would be taller....)

Om is going to be playing the role of Simon Cowell

Its that similar technology to Riya?

yep, but video focused.

Looked pretty good.

Marissa from Google just asked about their algorithms...........

Shirts are cool!   You guys need a flash on ur camera.

Too far away for flash.

Interesting.  The template extraction and the algorithm must be pretty complex to do this right.

Om pushing the 'how you going to grow' question?

And are they going to turn off the p2p?

First panel;

Are they using the video search technology on P2P networks?  Would be great to help spot fakes on BT.....

The p2p one was web search. Using your own browsing to actually cache the net.  Slowly builds an index.

Sort of a personal index of your stuff?

I wonder if the face regoc technology is the same one they use for airports etc

They said they've been working on it for 30 years, but this one seems very consumer focused. Like, search for Elle McPherson and it will find videos across multiple video indexes.

video existed 30 years ago ?:)

The concept is similar.  You scan a picture, the sw extracts a digital map from it (a template) which contains the unique features of the subject, and the matching algorithm searches other videos for a similar set of features.  It works like the biometric fingerprint recognition stuff but it's probably harder to do as you're trying to create and match from a 2D representation of a 3D subject.  It probably needs some assistance to refine the search manually.

They've been doing AI for 30 years.....

Needs a hefty CPU.  Fortunately CPU horsepower is cheap, unlike when I was working in the biometric space 9 years ago.  Try going all that with a 486 core....

wow that is cool, how they can pick a face out of a picture

well... even my $250 little pocket dig camera has face recognition

funny enough, even the little square brackets around the face looks the same

of course, my camera can't tell me who's face i'm looking at

Wow, that looks great.

More!

Showing

more

demos!

where's our man on the floor ?

hopefully in session 3

he's back!

I''m now with John from England from http://www.itema.com

Hey John, you came a long way for the conference!

Seems interesting.  Like the single sign-on/storing of credentials for insertion into different apps.

Some bold integrations ITema is offering

Cool idea as I hate typing the same stuff all the time.  Big barrier to adoption of disparate apps, IMHO.

The integration prospects look great for itema and everyone like easy installing of apps!

Viewdle vid: