Facebook (NASDAQ:FB) is trying to close the gap between humans and computers in facial recognition as the company says it has developed a technology that recognizes whether two different images are displaying the same face -- an ability that comes very close to replicating human ability to make the distinction. The new technology, called DeepFace, is claimed to be 97.25 percent accurate, reducing the margin of error with current state-of-the-art technology by more than 25 percent. According to Facebook, DeepFace is closely approaching human-level performance, which has scored 97.5 percent in the same standardized test. "In modern face recognition, the conventional pipeline consists of four stages: detect align represent classify," Facebook said in a research paper, released last week. "We revisit both the alignment step and the representation step by employing explicit 3D face modeling in order to apply a piecewise affine transformation, and derive a face representation from a nine-layer deep neural network."

The problem is that it is fairly easy to create things that behave like neurons, the brains major component. What is not easy is working out what the whole thing does once you have assembled it. It is assumed that neurons get excited by other neurons and when they get excited enough they "fire" and send their excitement on to other connected neurons. This is very easy to model but how do you determine how the neurons should be connected and what should govern the strengths of connection?

Now it turns out that they probably did work all along but we weren't doing things in quite the right way and we had no clear idea of the scale needed. To make neural networks fulfill their promise you need to first give then some deep structure and not rely on a random or simplistic architecture. Next you need to train big systems with big data - lots of it. Until quite recently finding enough data in the right form, and finding the large amounts of computer power to do the training, was a difficult problem. The data problem has been eased by the growth of the web and the computing problem by the growth of cloud computing.

IBM is big on cognitive. The company's recent AlchemyAPI acquisition is only the latest of many moves in the space. This particular acquisition adds market-proven text and image processing, backed by deep learning, a form of machine learning that resolves features at varying scales, to the IBM Watson technology stack. But IBM is by no means the only company applying machine learning to natural language understanding, and it's not the only company operating under the cognitive computing banner.

And yet, although AI has conquered many of the high castles of human intellect it is still limited because it lacks our ability for general reasoning. AI systems can do any the above 5 things better than any human, but there is not a single AI that can do all 5 things together, or more. "General intelligence" remains the Holy Grail for AI research. Once achieved, we will have arrived at the beginning of a truly intelligent mechanical mind. Nevertheless, DeepMind's seminal paper last year in Nature demonstrated how AI could develop general intelligence; in the example presented in the paper a deep learning algorithm was able to play many different Atari games by reasoning from first principles.

But lately, it has been aiming much more directly at building HAL, or what's sometimes called the Google Brain. As I wrote in a recent article, a fast-emerging branch of artificial intelligence called deep learning is helping Google and other companies and researchers produce significant advances in machines that at least approach the way we think. It won't be long--for better or worse--before their work also has a profound impact on marketing and advertising as well. Deep-learning software tries to emulate, albeit in a still primitive way, the activity in layers of neurons in the neocortex, the part of the human brain where thinking occurs. The software, part of artificial neural networks that simulate the neocortex's large array of neurons, learns quite literally to recognize patterns in digital representations of sounds, images, and other data.

When I arrived at a Stanford University auditorium Tuesday night for what I thought would be a pretty nerdy panel on deep learning, a fast-growing branch of artificial intelligence, I figured I must be in the wrong place--maybe a different event for all the new Stanford students and their parents visiting the campus. Despite the highly technical nature of deep learning, some 600 people had shown up for the sold-out AI event, presented by VLAB, a Stanford-based chapter of the MIT Enterprise Forum. The turnout was a stark sign of the rising popularity of deep learning, an approach to AI that tries to mimic the activity of the brain in so-called neural networks. In just the last couple of years, deep learning software from giants like, Facebook, and China's Baidu as well as a raft of startups, has led to big advances in image and speech recognition, medical diagnostics, stock trading, and more. "There's quite a bit of excitement in this area," panel moderator Steve Jurvetson, a partner with the venture firm DFJ, said with uncustomary understatement.

A computer that taught itself to play almost 50 video games including Space Invaders and Pong is being hailed as the pinnacle of artificial intelligence. But it is unlikely to spark the Terminator-like Armageddon predicted in recent months by technology entrepreneur Elon Musk (who provided early funding for the project) and physicist Stephen Hawking. Despite mastering more than half the classic Atari 2600 games, the program – deep Q-network (DQN), developed by DeepMind Technologies – struggled with more difficult challenges, such as, well, Pac-Man. "On the face of it, it looks trivial in the sense that these are games from the '80s and you can write solutions to them quite easily," said Dr Demis Hassabis, the vice-president of engineering at DeepMind, a British company acquired by a year ago for a reported £400m (US$650m). Never before has a computer taught itself how to do a range of complex operations, said Dr Hassabis, one of the company's co-founders.

While most would point to home security cameras as the primary application for imaging in the smart home - just this week, after all, smart home darling Nest launched their own home security cam - there appears to be a new focus in the connected home when it comes to imaging tech: our food. Just consider: Last month it was revealed by Science that had been doing research into machine learning around food identification, and had released a new app called Im2Calories, which examines an image and attempts to quantify the amount of calories on a plate. It uses "deep learning" technology - essentially a form of machine learning. Im2Calories can draw connections between what a given piece of food looks like, and vast amounts of available caloric data." And while we're used to Google doing crazy bleeding edge stuff, they're definitely not the only ones who see cameras as a natural fit in the kitchen. Last week we learned of a new product called the June Intelligent Oven, which uses images captured from an in-oven camera to identify food and then automatically program cooking time and temperature. And then there's the SmartPlate, a new product currently on Kickstarter from Fitly that includes three cameras in the plate itself. The cameras are used to detect food quantity and type the image across a database of food and associated caloric content. Wait, a plate with cameras? How exactly does that work? CEO Anthony Ortiz told me that the cameras will be recessed within the plate on the rim. "Think about the cameras having lenses .

Here's an odd fact: Turn-of-the-century photographers used to tell subjects to say "prunes" rather than "cheese," so that they would smile less. By studying nearly 38,000 high-school yearbook photos taken since 1905, UC Berkeley researchers have shown just how much smiling, fashion and hairstyles have changed over the years. The goal was not just to track trends, but figure out how to apply modern data-mining techniques and machine learning to a much older medium: photographs. Their research could advance deep-learning algorithms for dating historical photos and help historians study how social norms change over time. The main challenge for the team was to collect enough photos to create an "average" student profile for each decade from the 1900s to the 2010s.