Artificial intelligence (AI) is an exciting field of computer science. New advances in AI science bring closer the breakthrough moment when the first machine will outsmart humans. Deep learning, an AI revolutionary technique, helps machines to learn better. The technique consists of feeding huge data sets of annotated examples to the machines.
A hot area studied by computer science today is natural language processing. Scientists aim to teach AI machines how to read a document, understand its content, and answer questions about it.
In order to achieve this goal, AI machines using deep learning technique are fed with data carefully annotated for the purpose from huge databases. Until today, these databases were not large enough to be useful for this purpose.
A breakthrough was possible today, thanks to the work at Google DeepMind in London. Karl Moritz Hermann together with a team of computer scientists found their inspiration in the special way that CNN and Daily Mail write their online news articles.
Two other breakthroughs in computer science made possible by the 'deep learning revolution'. The first one comes from the field of neural networks. New techniques were developed by computer scientists to train networks with many layers. "Read-made" networks that are ready to learn are created this way. The second breakthrough consists of creating large databases annotated carefully for computers to learn from.
The task of creating a written word annotated database is difficult. For such a database the annotation must describe the content of the text. Then a learning algorithm is trying to understand the link by looking at the casual relationships and grammatical links, beyond the mere occurrence of words and phrases.
The large volume of article that are already available online creates a database that can be used by AI computers to learn and answer questions related to the material found in the online database. In other words, CNN and Daily Mail articles are used by DeepMind to teach machines how to read. The annotated database is easy to create by taking the bullet point summaries as the annotations and the news articles as the texts.
This revolutionary breakthrough at Google DeepMind sets the scene for more fascinating developments in the near future. Soon, Artificial Intelligence computers capable of reading will become just a banality. It is estimated by experts in Artificial Intelligence that deep learning will affect all the fields of activity, from video games, facial and voice recognition, to economics and fashion.