Fei-Fei Li: How we're teaching computers to understand pictures

TED 2015
Fei-Fei Li: How we're teaching computers to understand pictures (script)
Summary
The speaker is a computer vision expert. She and her team finally produced one of the first computer models. It can describe a picture when it sees it for the first time like a little girl who will be around three years old saw the picture and talk about it. It means that the computer can see and understand what it is and generate sentences to describe something that it sees. It's like the human brain integrates vision and language, they developed a model that connects parts of visual with words and phrases in sentences.
First, they used the computer algorithm in a mathematical language, but the computer can't see anything that is hidden or change a pose. Next, they used a big data that is enormous quantity and quality of pictures. Now, there are the great algorithms that are capable learning called convolutional neural network. By giving those data to the algorithms, the computer started learning like a child learns. The data became to nourish computer brain.
It took them a long time and the data they delivered was an unprecedented scale. And then, the computer could work like getting the brain which is age three. It makes some mistakes still, however, the machines were given sight. The computer has had visual intelligence. If the human continues to teach them, they can see brtter and they can help people for a better future.
Words in this story
integrate / combine, amalgamate
vision /  sight, sense of sight.   visual / graphic, image.
nourish / cultivate, feed, provide for.
specific / particular, special