Ever given a thought to how without applying much effort, we do things and don’t pay attention to them just because it’s simple. Well guess what the same things may be really tough for a computer to “understand” the way we do. The human mind is so complex and capable that we learn things by looking at the procedure of doing something at most twice in general. But, computers are not capable of remembering how the letter A is made without feeding it hundreds or even thousands of samples before it “learns” to recognize the letter. Not until now.
Scientists have taken inspiration from themselves, well more precisely, the human brain and the way it learns things in daily life and have come out with an AI (Artificial Intelligent) software that is capable of picking up new knowledge in a much more sophisticated and efficient way. The AI is so efficient that it recognizes a handwritten character about as accurately as a human can. Even deep learning which is the best existing machine learning technique, requires thousands of samples to distinguish between two significant characters drawn by human hands.
ALSO READ : Future robots will be controlled by thoughts
This type of learning will have a major advantage as it will require less data to work. The present machine learning algorithms that do terrific jobs like facial recognition, driving cars, speech recognition and whatnot require tons of data to work accurately and precisely. The new learning algorithm developed by Brenden Lake, a researcher at New York University, together with Ruslan Salakhutdinov, an assistant professor of computer science at the University of Toronto, and Joshua Tenenbaum, a professor in the Department of Brain and Cognitive Sciences at MIT, will prove to be a major step in more efficient machine learning and may prove to be a standard for researchers to dive into more artificial intelligence related applications.
The researchers used a learning technique called Bayesian program learning (BPL) framework. The software works by generating a unique program for every character using strokes of an imaginary pen. The program is matched to the particular character using a probabilistic programming technique. A new program is generated for an unfamiliar character. The software learns to recognize and recreate characters as the adults do.
When the researchers tested their model, less than 25 percent of the judges were able to tell the difference between the human – written and machine – made characters which seems really impressive. The research will help in many ways like learning of new words in the spoken language by the computer or even your smartphone. This could pave the way for machines actually understanding us in a more human – way and not just simply doing a translation.