Ray O’Farrell (@ray_ofarrell) takes the stage and mentions that the Thursday session focuses on innovation. With that Ray introduces Dr. Fei-Fei Li (@drfeifei) who is the director of the Stanford research labs. The Dr. contrasts the capabilities of a three year child who recognizes the world around them with our current compute devices. While a child can comprehend things around them, even our most powerful computers lack these basic skill sets. While computers can take incredibly detailed pictures, they cannot “see” the world around them.
Dr. Fei-Fei has led the Computer Vision and General Learning Lab at Stanford. The goal is for computers to see an object and understand what is happening around it. To do this they have to train the computer through a series of images to understand what an object is.
Dr. Fei-Fei’s team used images from the internet through a project called Imaginet which categorized billions of images. It took 50,000 people working across 167 companies. They then opened up the images through www.image-net.org project. Now that they had all this big data they could leverage the Convulotional Neural Network algorithms to recognize objects.
Using this algorithm they were able to learn things by analyzing google street view images. The algorithm was able to see patterns and define objects in cluttered images. The next step is to connect words and phrases based on snippets of the image to form sentences such as “the cat is lying on the bed”.
Think of a world where Dr’s can have a tireless set of eyes to monitor patient, robots that can search disaster zones and travel and explore new worlds.