A new program improves Artificial intelligence to the point where the machine no longer requires deep learning processes, that is, exhaustive repetitions to acquire new information. This new learning technique is completely different than previously used face and speech recognition systems and it gives hope to researchers that computers can become more humane.
The new program that was developed with the help of researchers from the Massachusetts Institute of Technology and the University of Toronto relies on a new technique called the Bayesian Program Learning.
This model presupposes that computers use inference to learn new data, whereas former programs had to be repeatedly exposed to a piece of information to learn it. The latter required too much time on behalf of computers and they were not very effective, either, scientists have explained.
To test the abilities of the new program, scientists gave the Artificial Intelligence machinery a set of handwritten symbols and asked it to reproduce them. Surprisingly, the computer was able to recognize and re-create a letter even if it was exposed to it only once.
Researchers have explained that the model was much faster because it used a system of inferences based on the strokes of pen that the writer applied. The computer immediately learned which moves the writer was supposed to make and, thus, was able to memorize the letters in a shorter period of time.
The good news is: it is not only the learning process that has been shortened and simplified, but the entire AI program has become more humane due to the new changes. Scientists recon that computers will become more like human beings if they use logical inference instead of repetition. This is, actually how humans acquire new knowledge, the moment they are exposed to new data.
Now that the efficiency of the new Bayesian program has been proven, researchers aim to improve existing face recognition and speech recognition technology. The new systems will eventually lead to the development of an AI computer that can actually understand language, or so scientists hope.
Image source: www.pixabay.com