Coupled with artificial intelligence, deep learning has been making headlines recently with tech giants such as Google and Facebook turning to the technology to come up with futuristic solutions.

Some have expressed concerns about computers ultimately outsmarting and dominating humans, as in many science-fiction scenarios. But as the saying by Chinese military strategist Sun Tzu goes, if you know your enemies and know yourself, you will not be imperiled in a hundred battles.

So, with that in mind, what is the big fuss about deep learning?

The concept of deep-learning technology was first introduced by Kunihiko Fukushima in 1980.

Deep learning basically aims to create computer systems that can better understand complex data such as pictures, voices, and handwriting. It is closely related with neural networks and machine learning, two buzzwords in the tech industry.

A neural network is a type of computer architecture consisting of network nodes that exchange messages with each other. Upon receiving signals, each node conducts an independent task and exchanges information with other node peers to achieve the same goal, or goals, let’s say for instance, to tell a cat apart from a dog.

Deep learning is an advanced type of machine learning with a set of algorithms and multiple layers of network nodes. It can reinforce the recognition capabilities of neural network systems through a sort of self-training, for instance, watching millions of pictures of cats and dogs.

Although deep-learning technology has made great strides in the past decades, Seth Lloyd, professor of quantum mechanical engineering at Massachusetts Institute of Technology, said that the intelligence of current deep-learning technology is still too narrow and brittle, compared to that of a human.

“Deep learning is informationally broad -- it analyzes vast amounts of data -- but conceptually shallow,” said the MIT professor in an article, arguing that there are certain limits in building machines that can learn and think like a human.

Despite such limitations, many tech companies already utilize deep-learning technology. There are even plans to take bold steps to deploy the technology in a wide range of sectors including health care and environment.

Google’s deep-learning program PlaNet can locate places just by looking at pictures.

DeepMind, Google’s affiliate and developer of the AI program AlphaGo, aims to tackle some of the toughest global issues, such as incurable diseases and global warming.

“Our hope is that one day they (AIs) could be extended to help us address some of society’s toughest and most pressing problems, from climate modelling to complex disease analysis,” said Demis Hassabis, the chief of the Google-owned start-up.

IBM, one of the pioneers in the AI sector, has been running a series of projects utilizing AI techniques.

The company’s Deep Blue chess-playing computer came to the spotlight in 1997 when it became the first computer system to defeat a world champion in chess. Its AI system Watson won the Jeopardy quiz game show in 2011.

IBM has been putting much weight on developing what it calls “cognitive computing,” which are solutions that understand, reason, and learn just like humans.

“It is the dawn of a new era, the cognitive era, digital business plus digital intelligence,” said IBM CEO Ginni Rometty at the CES trade show in Las Vegas in January.

In partnership with firms such as fitness company Under Armor, medical tech firm Medtronics, and Internet firm Softbank, IBM is trying to incorporate its AI systems into products and services.

Watson currently serves as a training coach for users of the Under Armor fitness app. It collects and analyzes users’ data. It also warns diabetics who use Medtronics devices of low blood sugar levels and works as a brain that powers Softbank’s Pepper humanoid robot.

Watson offers more than 30 services, including voice recognition and computer vision.

A wide usage of AI programs across industries is also expected to benefit graphics chip manufacturers, such as Nvidia, since the computer programs need powerful brains to make accurate decisions and predictions based on images.

By Kim Young-won (wone0102@heraldcorp.com)