EFEVioleta Molina Gallardo Madrid

Recent advances in artificial intelligence (AI) have endowed certain machines with human qualities such as intuition and improvisation.

Despite such technological progress, experts say future machines may never acheive self-awareness.

"We don't really understand how consciousness works, we barely know anything at all about this human process," Greg Corrado, the neuroscientist behind Google's AI project Google Brain, told EFE on Tuesday.

"We will probably never be able to build machines that are self-aware," he added.

After 60 years of research and a decade of significant advances in AI, there are computers capable of understanding basic concepts about the world and engaging in human communication, Corrado explained.

These machines, which are being developed by tech companies and prestigious universities, may be able to perform highly-specialized tasks with precision and efficacy, but they remain light-years away from developing consciousness.

Although they can learn from experience, these latest high-tech contraptions are completely unable to possess inherent human qualities such as common sense, consciousness, or social and creative skills.

Corrado emphasized the fact that the inner workings of the human brain and its subjective experience are still, for the most part, scientific mysteries.

The aim of AI, he added, is not to replicate human intelligence (although AI researchers often draw from the rich well of neuroscience), but to create "seemingly intelligent" systems.

Systems that can now write original poems due to their previous knowledge of literary classics; systems that are able to beat world champions at the traditional Chinese strategy game known as "go," systems that detect the typologies of objects that appear in a picture: at very specific tasks, they are much better than humans.

Corrado underscored the latest breakthroughs in "deep learning"; that is, machines that learn on their own through examples and are able to perform certain tasks based on a set of algorithms that model abstract data.

For example, a system could deduce the defining characteristics of cats after scanning through a large number of cat pictures, and later identify a feline in a completely new image.

Perceptual computational systems, which are able to "see," "hear" and "understand" the physical world, are seeing great leaps in progress, said Corrado.

There are some that are now capable of detecting patterns while lacking all the needed information, which, according to Corrado, makes them somewhat "intuitive."

In the field of "deep learning," artificial neural networks are being created that are inspired by information-processing in human nervous systems.

Other machines possess a certain degree of improvisational abilities: for example, AlphaGo, the computer that beat the "go" world champion, won due to "surprising and improvised" moves after interpolating and extrapolating its available data.

Corrado, however, cautioned that these extraordinary feats do not equate to creativity.

Although deep learning is starting to be used in the arts as a creative tool _ for example, to discover new musical combinations _ the machines still need information supplied by humans to function.

Will AI become self-aware by itself, without any human intervention?

"We don't actually know how the human imagination works, nor do we understand the knowledge processes," Corrado said.

In his opinion, artists are heavily influenced by their social contexts and their eagerness to communicate with one another, both traits that machines lack.

"Whatever machines create, it will be something very different from what we consider to be the fine arts," Corrado added.

vmg/dl/hh