Derek Schloss

“Deep learning requires much more data than humans, but once trained on big data, it will outperform humans by far for a given task, especially in dealing with quantitative optimization (like picking an ad to maximize likelihood of purchase, or recognizing a face out of a million possible faces). While humans are limited in the number of things they can pay attention to at once, a deep-learning algorithm trained on an ocean of information will discover correlations between obscure features of the data that are too subtle or complex for we humans to comprehend, and which may not even be noticed.”
― Kai-Fu Lee, AI 2041: Ten Visions for Our Future

“When we “see,” we are actually applying our accumulated knowledge of the world—everything we’ve learned in our lives about perspective, geometry, common sense, and what we have seen previously. These come naturally to us but are very difficult to teach a computer. Computer vision is the field of study that tries to overcome these difficulties to get computers to see and understand. COMPUTER VISION APPLICATIONS We are already using computer vision technologies every day. Computer vision can be used in real time, in areas ranging from transportation to security. Existing examples include: driver assistants installed in some cars that can detect a driver who nods off autonomous stores like Amazon Go, where cameras recognize when you’ve put a product in your shopping cart airport security (counting people, recognizing terrorists)”
― Kai-Fu Lee, AI 2041: Ten Visions for Our Future

“In order to construct a flawless imitation, the first step was to gather as much video data as possible with a web crawler. His ideal targets were fashionable Yoruba girls, with their brightly colored V-neck buba and iro that wrapped around their waists, hair bundled up in gele. Preferably, their videos were taken in their bedrooms with bright, stable lighting, their expressions vivid and exaggerated, so that AI could extract as many still-frame images as possible. The object data set was paired with another set of Amaka’s own face under different lighting, from multiple angles and with alternative expressions, automatically generated by his smartstream. Then, he uploaded both data sets to the cloud and got to work with a hyper-generative adversarial network. A few hours or days later, the result was a DeepMask model. By applying this “mask,” woven from algorithms, to videos, he could become the girl he had created from bits, and to the naked eye, his fake was indistinguishable from the real thing. If his Internet speed allowed, he could also swap faces in real time to spice up the fun. Of course, more fun meant more work. For real-time deception to work, he had to simultaneously translate English or Igbo into Yoruba, and use transVoice to imitate the voice of a Yoruba girl and a lip sync open-source toolkit to generate corresponding lip movement. If the person on the other end of the chat had paid for a high-quality anti-fake detector, however, the app might automatically detect anomalies in the video, marking them with red translucent square warnings”
― Kai-Fu Lee, AI 2041: Ten Visions for Our Future

“Many people assume AI is “programmed” or “taught” by humans with specific rules and actions, like “cats have pointy ears and whiskers.” But deep learning actually works better without these external human rules. Instead of being nudged by humans, many examples of a given phenomenon are fed into the input layer of a deep learning system, along with the “correct answer” at the output layer. In this way, the network in between the input and output can be “trained” to maximize the chance of getting the correct answer to a given input.”
― Kai-Fu Lee, AI 2041: Ten Visions for Our Future