Peter Fingar: The Cognitive Computing Era is Upon Us
Business-tech author details how cognitive systems wield a power like we've never seen and could soon be tending to our agriculture, industry, and service needs
The era of cognitive systems is dawning and building on today’s computer programming era. All machines, for now, require programming, and by definition programming does not allow for alternate scenarios that have not been programmed. To allow alternating outcomes would require going up a level, creating a self-learning Artificial Intelligence (AI) system. Via biomimicry and neuroscience, cognitive computing does this, taking computing concepts to a whole new level.
Fast forward to 2011 when IBM’s Watson won Jeopardy! Google recently made a $500 million acquisition of DeepMind. Facebook recently hired NYU professor Yann LeCun, a respected pioneer in AI. Microsoft has more than 65 PhD-level researchers working on deep learning. China’s Baidu search company hired Stanford University’s AI Professor Andrew Ng. All this has a lot of people talking about deep learning. While artificial intelligence has been around for years (John McCarthy coined the term in 1955), “deep learning” is now considered cutting-edge AI that represents an evolution over primitive neural networks.
Taking a step back to set the foundation for this discussion, let me review a few of these terms.
As human beings, we have complex neural networks in our brains that allow most of us to master rudimentary language and motor skills within the first 24 months of our lives with only minimal guidance from our caregivers. Our senses provide the data to our brains that allows this learning to take place. As we become adults, our learning capacity grows while the speed at which we learn decreases. We have learned to adapt to this limitation by creating assistive machines. For over 100 years machines have been programmed with instructions for tabulating and calculating to assist us with better speed and accuracy. Today, machines can be taught to learn much faster than humans, such as in the field of machine learning, that can learn from data (much like we humans do). This learning takes place in Artificial Neural Networks that are designed based on studies of the human neurological and sensory systems. Artificial neural nets make computations based on inputted data, then adapt and learn. In machine learning research, when high-level data abstraction meets non-linear processes it is said to be engaged in deep learning, the prime directive of current advances in AI. Cognitive computing, or self-learning AI, combines the best of human and machine learning and essentially augments us.
When we associate names with current computer technology, no doubt Steve Jobs or Bill Gates come to mind. But the new name will likely be a guy from the University of Toronto, the hotbed of deep learning scientists. Meet Geoffrey Everest Hinton, great-great-grandson of George Boole, the guy who gave us the mathematics that underpin computers.
Hinton is a British-born computer scientist and psychologist, most noted for his work on artificial neural networks. He is now working for Google part-time, joining AI pioneer and futurist Ray Kurzweil, and Andrew Ng, the Stanford University professor who set up Google’s neural network team in 2011. He is the co-inventor of the back propagation, the Boltzmann machine, and contrastive divergence training algorithms and is an important figure in the deep learning movement. Hinton’s research has implications for areas such as speech recognition, computer vision and language understanding. Unlike past neural networks, newer ones can have many layers and are called “deep neural networks.”
As reported in Wired magazine, “In Hinton’s world, a neural network is essentially software that operates at multiple levels. He and his cohorts build artificial neurons from interconnected layers of software modeled after the columns of neurons you find in the brain’s cortex—the part of the brain that deals with complex tasks like vision and language.
“These artificial neural nets can gather information, and they can react to it. They can build up an understanding of what something looks or sounds like. They’re getting better at determining what a group of words mean when you put them together. And they can do all that without asking a human to provide labels for objects and ideas and words, as is often the case with traditional machine learning tools.
“As far as artificial intelligence goes, these neural nets are fast, nimble, and efficient. They scale extremely well across a growing number of machines, able to tackle more and more complex tasks as time goes on. And they’re about 30 years in the making.”
How Did We Get Here?
Back in the early 80s, when Hinton and his colleagues first started work on this idea, computers weren’t fast or powerful enough to process the enormous collections of data that neural nets require. Their success was limited, and the AI community turned its back on them, working to find shortcuts to brain-like behavior rather than trying to mimic the operation of the brain.
But a few resolute researchers carried on. According to Hinton and Yann LeCun (NYU professor and Director of Facebook’s new AI Lab), it was rough going. Even as late as 2004—more than 20 years after Hinton and LeCun first developed the “back-propagation” algorithms that seeded their work on neural networks—the rest of the academic world was largely uninterested.
By the middle aughts, they had the computing power they needed to realize many of their earlier ideas. As they came together for regular workshops, their research accelerated. They built more powerful deep learning algorithms that operated on much larger datasets. By the middle of the decade, they were winning global AI competitions. And by the beginning of the current decade, the giants of the Web began to notice.
Deep learning is now mainstream. “We ceased to be the lunatic fringe,” Hinton says. “We’re now the lunatic core.” Perhaps a key turning point was in 2004 when Hinton founded the Neural Computation and Adaptive Perception (NCAP) program (a consortium of computer scientists, psychologists, neuroscientists, physicists, biologists and electrical engineers) through funding provided by the Canadian Institute for Advanced Research (CIFAR).
Back in the 1980s, the AI market turned out to be something of a graveyard for overblown technology hopes. Computerworld’s Lamont Wood reported, “For decades the field of artificial intelligence (AI) experienced two seasons: recurring springs, in which hype-fueled expectations were high; and subsequent winters, after the promises of spring could not be met and disappointed investors turned away. But now real progress is being made, and it’s being made in the absence of hype. In fact, some of the chief practitioners won’t even talk about what they are doing.
But wait! 2011 ushers in a sizzling renaissance for A.I.
How did we get here? What’s really new in A.I.?
Let’s touch on some of these breakthroughs.
What’s really, really new? Deep Learning.
Machines learn on their own? Watch this simple everyday explanation by Demis Hassabis, cofounder of DeepMind.
It may sound like fiction and rather far-fetched, but success has already been achieved in certain areas using deep learning, such as image processing (Facebook’s DeepFace) and voice recognition (IBM’s Watson, Apple’s Siri, Google’s Now and Waze, Microsoft’s Cortana and Azure Machine Learning Platform).
Beyond the usual big tech company suspects, newcomers in the field of Deep Learning are emerging: Ersatz Labs, BigML, SkyTree, Digital Reasoning, Saffron Technologies, Palantir Technologies, Wise.io, declara, Expect Labs, BlabPredicts, Skymind, Blix, Cognitive Scale, Compsim’s (KEEL), Kayak, Sentient Technologies, Scaled Inference, Kensho, Nara Logics, Context Relevant, Expect Labs, and Deeplearning4j. Some of these newcomers specialize in using cognitive computing to tap Dark Data, a.k.a. Dusty Data, which is a type of unstructured, untagged and untapped data that is found in data repositories and has not been analyzed or processed. It is similar to big data but differs in how it is mostly neglected by business and IT administrators in terms of its value.
Machine reading capabilities have a lot to do with unlocking “dark” data. Dark data is data that is found in log files and data archives stored within large enterprise class data storage locations. It includes all data objects and types that have yet to be analyzed for any business or competitive intelligence or aid in business decision making. Typically, dark data is complex to analyze and stored in locations where analysis is difficult. The overall process can be costly. It also can include data objects that have not been seized by the enterprise or data that are external to the organization, such as data stored by partners or customers. IDC, a research firm, stated that up to 90 percent of big data is dark.
Cognitive Computing uses hundreds of analytics that provide it with capabilities such as natural language processing, text analysis, and knowledge representation and reasoning to …
- make sense of huge amounts of complex information in split seconds,
- rank answers (hypotheses) based on evidence and confidence, and learn from its mistakes.
DeepQA technology, and continuing research underpinning IBM’s Watson, is aimed at exploring how advancing and integrating Natural Language Processing (NLP), Information Retrieval (IR), Machine Learning (ML), Knowledge Representation and Reasoning (KR&R) and massively parallel computation can advance the science and application of automatic Question Answering and general natural language understanding.
Cognitive computing systems get better over time as they build knowledge and learn a domain—its language and terminology, its processes and its preferred methods of interacting.
Unlike expert systems of the past that required rules to be hard coded into a system by a human expert, cognitive computing systems can process natural language and unstructured data and learn by experience, much in the same way humans do. As far as huge amounts of complex information (Big Data) is concerned, Virginia “Ginni” Rometty, CEO of IBM stated, “We will look back on this time and look at data as a natural resource that powered the 21st century, just as you look back at hydrocarbons as powering the 19th.”
And, of course, this capability is deployed in the Cloud and made available as a cognitive service, Cognition as a Service (CaaS).
With technologies that respond to voice queries, even those without a smartphone can tap Cognition as a Service. Those with smart phones will no doubt have Cognitive Apps. This means 4.5 billion people can contribute to knowledge and combinatorial innovation, as well as the GPS capabilities of those phones to provide real-time reporting and fully informed decision making: whether for good or evil.
Geoffrey Hinton, the “godfather” of deep learning, and co-inventor of the back propagation and contrastive divergence training algorithms has revolutionized language understanding and language translation. A pretty spectacular December 2012 live demonstration of instant English-to-Chinese voice recognition and translation by Microsoft Research chief Rick Rashid was one of many things made possible by Hinton’s work. Rashid demonstrates a speech recognition breakthrough via machine translation that converts his spoken English words into computer-generated Chinese language. The breakthrough is patterned after deep neural networks and significantly reduces errors in spoken as well as written translation. Watch:
Artificial General Intelligence
According to the AGI Society, “Artificial General Intelligence (AGI) is an emerging field aiming at the building of ‘thinking machines;’ that is, general-purpose systems with intelligence comparable to that of the human mind (and perhaps ultimately well beyond human general intelligence). While this was the original goal of Artificial Intelligence (AI), the mainstream of AI research has turned toward domain-dependent and problem-specific solutions; therefore it has become necessary to use a new name to indicate research that still pursues the ‘Grand AI Dream.’ Similar labels for this kind of research include ‘Strong AI,’ ‘Human-level AI,’ etc.”
AGI is associated with traits such as consciousness, sentience, sapience, and self-awareness observed in living beings. “Some references emphasize a distinction between strong AI and ‘applied AI’ (also called ‘narrow AI’ or ‘weak AI’): the use of software to study or accomplish specific problem solving or reasoning tasks. Weak AI, in contrast to strong AI, does not attempt to simulate the full range of human cognitive abilities.”
Turing test? The latest is a computer program named Eugene Goostman, a chatbot that “claims” to have met the challenge, convincing more than 33 percent of the judges at this year’s competition that ‘Eugene’ was actually a 13-year-old boy.
The test is controversial because of the tendency to attribute human characteristics to what is often a very simple algorithm. This is unfortunate because chatbots are easy to trip up if the interrogator is even slightly suspicious. Chatbots have difficulty with follow up questions and are easily thrown by non-sequiturs that a human could either give a straight answer to or respond to by specifically asking what the heck you’re talking about, then replying in context to the answer. Although skeptics tore apart the assertion that Eugene actually passed the Turing test, it’s true that as AI progresses, we’ll be forced to think at least twice when meeting “people” online.
Isaac Asimov, a biochemistry professor and writer of acclaimed science fiction, described Marvin Minsky as one of only two people he would admit were more intelligent than he was, the other being Carl Sagan. Minsky, one of the pioneering computer scientists in artificial intelligence, related emotions to the broader issues of machine intelligence, stating in his book, The Emotion Machine, that emotion is “not especially different from the processes that we call ‘thinking.’”
Considered as one of his major contributions, Asimov introduced the Three Laws of Robotics in his 1942 short story “Runaround,” although they had been foreshadowed in a few earlier stories. The Three Laws are:
- A robot may not injure a human being or, through inaction, allow a human being to come to harm.
- A robot must obey the orders given to it by human beings, except where such orders would conflict with the First Law.
- A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
What would Asimov have thought had he met the really smart VIKI? In the movie, iRobot, V.I.K.I (Virtual Interactive Kinetic Intelligence) is the supercomputer, the central positronic brain of U. S. Robotics headquarters, a robotic distributor based in Chicago. VIKI can be thought of as a main