Researchers want to create computers that will be able to think like humans

Researchers are closer to when they will be able to build a computer that will be able to think like people, Reuters informs.

Because artificial intelligence and intelligent machines a reality, the future computers will be able to think like people considered experts in the field. Scientists have recently conducted a series of important advances towards this goal.

Scientists Americans announced Thursday that they have created a computer model – or algorithm – which holds that human ability unique to extract and understand the concepts of a single example, in a study that examined the learning of letters alphabet, handwritten unknown that software.

Studies of this kind shall set a double objective: to better understand the process of human learning and develop learning algorithms us closer to the human, said Brenden Lake, professor of cognitive science and analysis at New York University.

The new algorithm was designed to cause a computer to be able to learn quickly in a single instance, as people usually do.

“You look even a very young child a horse or a school bus or a skateboard, and they understand those images after only one or very few examples,” said Joshua Tenenbaum, a professor of cognitive science in computing the prestigious MIT (Massachusetts Institute of Technology).

For machines that are designed to teach standard algorithms used to take tens, hundreds or even thousands of examples preparatory to yield similar results, said Professor Tenenbaum.

In this recent study, computer holding the new algorithm and human volunteers have been given a set of letters selected from a database of 1,600 handwritten letters, coming from 50 types of alphabet used worldwide. Among them was even a fictional alien alphabet, presented in animated TV series “Futurama”.

One of the tasks assigned to human volunteers and stated that those computers to replicate various letters after they have been presented in a single instance. A jury – made up of people – was then asked to identify letters that have been replicated by a computer. Jurors found that the letters reproduced computers so they could not be distinguished from the replicated virtual human volunteers.

Ruslan Salakhutdinov, professor of statistics and science of computer University Toronto, hopes the new study will help to achieve considerable progress in the creation of artificial intelligence by developing a new generation of intelligent machines, which “will have an intelligence similar the people “.

The approach used in this study could be applied in other learning processes to be used for intelligent machines, such as voice recognition and object recognition.

US researchers study was published in the journal Science.

Related