Google "Cognition" - The killer app that will take advantage of GLASS

Google is as much about AI as it is about Search. In fact AI is the prime driver of almost all its products and is embedded in every product of Google is some way. Some prime examples are Search, Instant, Now, Spam detection, Voice recognition, Translate, Goggles, Maps, Self-driving cars, personalised ads, face tracking, image recognition and most recently its Neural networks that learned what a "Cat" is by sifting through thousands of videos on Youtube.

The application of AI evolved from purely text based apps at first, then sound based apps, and then image based and finally to video based. Although Google developed the Goggles application for 3D object recognition, it was not successful because of limited use case and practicality on phones. There has been no practical way of gathering that data at the volumes required for machine learning. Till now. With GLASS that will change. Youtube and Streetview had this capability to a certain extent but with GLASS Google will for the first time have a full fledged capability to capture and recognize 3D real world objects.

This app will essentially be a dramatic improvement over the Goggles app by taking advantage of the flexibility in field of view provided by GLASS and can do to object recognition, what Google Voice did to the voice recognition capability and what Google Translate did to the translation capability by learning through real world usage. GLASS will create thousands of (if successful millions) of 3D data volunteers.

I believe that AI will not reach its true potential only trough text, sound and image based applications. It could evolve much more quickly and gain much more knowledge of the world through 3D object recognition. This will literally add a new dimension of knowledge representation to AI agents. Human brains evolved to become as intelligent as they are now, primarily by living in a dynamic 3D world which they had to understand very well to survive.

Some of the applications of object recognition would be: (simplistic to advanced in that order)

  1. Pull-up information at real time about real world objects/famous people in your field of view (will be very useful in malls and in general for advertisements)

  2. Labeling physical objects and tracking them (automatic search for lost items just by looking around or notify if you are leaving an object behind

  3. Ability to search the web with real world references (search for pet stores which have that breed of dog I saw yesterday at the park)

  4. A database/index for the real world

  5. Location recognition just by looking around without the need for GPS/Network and understanding 3D space (3D version of maps)

  6. Improve dramatically the performance of augumented reality applications and creation of virtual objects which can replicate real object behaviour (virtual books which you can hold and flip through, virtual balls you can throw and catch,...)

  7. Ability to develop prediction models for real world events/actions (eg. a simple application would be to learn that the sound "meow" implies that there is a cat somewhere around you or that if the leaves of the tree near you are fluttering vigorously, it implies that there is wind and it is strong. These models will be the precursors to logical thinking in machines and will lead to AGI (artificial general intelligence). This will also be of significant help for robotics which requires a lot of real-time prediction of its environment