Google Lens Can Understand What Your Smartphone Camera Capturing
Google’s spokesperson Sundar Pichai has said, “Google Lens is a set of vision based computing capabilities, that can understand what you looking at and help you take action based on that information.” Which is, in the future Google is planning to embed Google Lens AI in the smartphone and thus your smartphone can understand what your camera see and help you to take action in a meaningful way.
Google spokesperson has introduced, some real time examples of Google Lens feature. In the first image, if you point Google Lens onto a flower and it can tell what flower it is, therefore if you are seeing an unknown object or something you can’t remember its name or application just point out your phone to that object and thus Google Lens will let you know what it is. Now in the second image, if you are on a street or somewhere else, Google Lens knows where you are and what are you looking at if you point your smartphone camera onto restaurant or something else Google Lense will give you right information about what your smartphone camera see in a meaningful way and in the third image you can get the username and password of a wifi router by just pointing your phone at the wifi router.
Tensor Processing Unit (TPU)
For the successful completeness of Google Lense, Google is building the AI first data centers for the machine learning. Still, google is using AI first data centers for the voice recognition, automatic replaying of Gmails, Google image search and so on. Google is rethinking on their current machine learning and AI based computational architecture for building Google Lens. In the last year, Google has launched the Tensor Processing Unit (TPU) developed for machine learning and also been used in RankBrain which Google used to provide search results. A Tensor Processing Unit is capable of computing over 100 million images per day and 15 to 30 time faster and 30 to 80 time more power efficient than a central processing unit (CPU) or Graphical Processing Unit (GPU).
Google Cloud TPU
The above picture is the one cloud TPU board which is been used for optimizing training and inference of machine learning for AI era. The board uses four chips and each individual boards are capable of 10080 trillion floating point operations per second. Google spokesperson has said, “Cloud TPU’s are coming to Google Compute Engine ” and is the reason behind to name it as cloud TPU.