When you think about the "incredible" tasks that a computer can handle, the first thing that comes to mind is the most complex calculations in a short time or the analysis of huge amounts of data - something that you yourself can never solve yourself. Or, I recall Lee Sedol's recent defeat in go, a classic strategy game. AI's latest victories have been made possible in large part by deep learning, which is now opening up all the possibilities for AI and the people behind it.
But simple, day-to-day tasks that even a child in their right mind can do seem to undermine the functionality of AI systems: things like identifying what food is on your plate or identifying emotions on another person's face. These easy tasks for humans were impossible for machines. Up to this point.
Deep learning techniques have brought common sense to machines. In the past, programmers wrote complex algorithms that described everything down to the smallest detail. Such an explicit and deterministic algorithm is suitable when you are faced with a big, awkward computation. Deep learning frees AI from this kind of limitation, allows the system to learn from its mistakes, remember everything it has learned, interact with users for more information.
The deep learning revolution is happening in large part because big data is becoming available for learning. A human toddler can learn what it needs after a few tries, but the machine will take much longer. Deep learning relies on access to vast amounts of data, as AI machines must base their choices on probabilities and statistical significance. A mechanical replacement for intuition has not yet been invented.
Deep possibilities
Advances in deep learning have already dramatically improved voice search capabilities: Google replaced the Android speech system with a new deep learning-based system, and errors dropped to 25 percent overnight. Cameras using deep neural networks can now read aloud to people and understand sign language. Facebook boasts that its deep learning capabilities have made the platform accessible to blind users by learning how to describe photos.
In the coming years, both large tech companies and many startups will start using deep learning to create new products and services, and to modernize existing applications. New markets and businesses will sprout and drive innovation, services and products. Deep learning systems will improve and become more accessible and easier to use. The easier it is to use them, the more our interaction with technology will change.
Promotional video:
Aditya Singh, a partner at Foundation Capital, believes that the development of the deep learning operating system will democratize deep learning and drive the widespread adoption of practical AI. The result will be that people can solve their pressing problems, or something more significant, using deep learning. In this sense, AI can become an equalization mechanism, allowing people of any class and state to change the world.
ILYA KHEL