The rise of machines might still be far ahead, but computers are getting smarter, faster, and more creative. Next-generation computing is changing our private lives, as well as industries like healthcare, insurance, banking, finance, retail, and many, many others. And while David Kenny, the general manager of IBM Watson, believes that AI can only be as smart as the people teaching it, this is definitely not the case for the latest wave of cognitive technologies.
What is cognitive computing?
Different sources provide us with different definitions of cognitive computing.
Forbes proposes that cognitive computing is a definition that describes a mashup of cognitive science—the study of the human brain and how it functions — and computer science. IBM ― the pioneers of cognitive computing ― describe it as “systems that learn at scale, reason with purpose, and interact with humans naturally.” Finally, International Conference on Cognitive Computing defines cognitive computing as “technology platforms that are based on the scientific disciplines of artificial intelligence and signal processing”.
As is clear from all these definitions, cognitive computing is the next step in the development of artificial intelligence. Cognitive computing uses machine learning, mining, pattern recognition, natural language processing, expert systems, neural networks, deep learning, robotics, and other techniques to learn how to think like a human. It’s better than older computing technologies in that it learns to understand natural language, recognize objects in an image, and can achieve the level of creativity and prediction that older technologies can’t.
Most of the time, cognitive computing is used to help people make decisions in complex situations that require the analysis of a large amount of structured and unstructured data. Over time, cognitive systems become better and faster at processing data and identifying patterns. Based on their experience, they even learn to anticipate new problems and model possible solutions.
Where is cognitive computing used?
Businesses use cognitive computing to identify trends and patterns and make informed and data-based decisions, understand the human language and interact with customers and employees more naturally, assess risks in real-time, and anticipate problems in the future. Here are some examples of how cognitive computing is used in different industries.
Healthcare
There are already a lot of ways AI is used in healthcare. Cognitive computing processes a large amount of structured and unstructured data which is electronic health records: patient histories, diagnoses, and conditions, as well as data from medical journals. Then, it can make treatment and diagnostical recommendations. This way it doesn’t make a decision instead of the medical professional but helps them with the decision-making process.
Retail
Personalization has become paramount in the marketing and customer service of every retail business. Cognitive computing is used to analyze existing information about the customer, search through existing products, and send personalized recommendations to existing customers and leads.
Banking and finance
Banks also use cognitive computing to analyze unstructured data from different sources to decide what kind of offers they should provide to their customers, how they should communicate with the customers, and what they should market to them. Customer service chatbots, often used in banking and finance apps, are also powered by cognitive computing.
Logistics
Logistics networks have never been more complex. The latest technology is needed to find the best solutions and meet common logistics challenges. With the introduction of cyber records and cognitive computing, many common problems can be solved. These include freight costs that are too high, transit times that are too long, or not charging enough for freight.
What are the disadvantages of cognitive systems?
It’s not only the complexity of the systems that make them challenging and inapplicable in many industries today. Other disadvantages also play a role, and we should attend to them all the way through research, development, and implementation.
Safety and security
Data protection is the biggest challenge to all companies willing to use cognitive systems. Often, the data used for machine learning as well as other technologies is sensitive. For example, in healthcare, industries get a hold of tons of private patient information. In finance and banking, leaks and other security issues might lead to financial fraud. When working with cognitive systems, cybersecurity is paramount.
Slow and expensive adoption
Cognitive computing doesn’t just require extensive research and development ― they are also hard to adopt. Whatever industry you take, you need trained specialists to work with technology this advanced. This means you have to train staff and invest a lot in implementation. For smaller organizations, this might be an impossible task, which will lead to them losing to bigger corporations in a long term.
Environmental impact
We live in a world in which it’s impossible to ignore environmental issues and responsibility. Сognitive systems and neural networks consume a lot of power and have a sizable carbon footprint.
What is neuromorphic technology?
You might’ve heard about semiconductors. Very basically speaking, a semiconductor is a substance that has specific electrical properties that enable it to serve as a foundation for computers and other electronic devices. With the development of the latest autonomous vehicles, robots, drones, and other self-reliant machines experts say semiconductors soon won’t be enough. The alternative is neuromorphic computing.
Neuromorphic computing works in the following way: it mimics the human brain and nervous system by establishing what is known as spiking neural networks. In neuromorphic computing, spikes from individual electronic neurons activate other neurons down a cascading chain. In the same way, the brain sends and receives signals from biological neurons that spark or recognize movement and sensations in our bodies. Unlike traditional computing, neuromorphic computing doesn’t rely on rigid binary zero or one structure. Because of this difference, neuromorphic chips compute much more flexibly than traditional ones.
The goal of neuromorphic technology is to make computers think creatively, as well as recognize people or objects they’ve never seen, predict problems, and create solutions. It’s safe to say that we’re not there yet. Neuromorphic computing is at the stage of development and exploration and is not ready to be applied in the real world industries. However, what researchers have done so far is also exciting. For example, recently Intel developed a neuromorphic robot that can see and recognize unknown objects based on just one example, unlike models based on machine learning technologies that require instructions, time, and data to learn. Similarly, Intel and Cornell University showed off Loihi ― a technology that closely replicates how the brain « smells » something. It can detect 10 different odors. This is something that could be eventually used for airport security, smoke and carbon monoxide detection, and quality control in factories.
When will neuromoprhic technologies enter the market?
Neuromorphic technologies don’t seem to be just around the corner ― however, they aren’t as unreachable as it may seem. According to Gartner, traditional computing technologies won’t be enough by 2025 and this will force a shift to the implementation of neuromorphic technologies and other qualitatively different technologies. At the same time, Emergen Research predicts that the global neuromorphic processing market will reach $11.29 billion by 2027.
Most researchers agree that neuromorphic technologies won’t completely replace the traditional ones ― they will add to the existing computing and allow things that haven’t been possible before.
What are the challenges of neuromorphic technology?
It’s crucial to understand that neuromorphic technology is a whole different approach to computing and it clashes significantly with existing computing norms. Neuromorphic technology requires different programming languages ― they will have to be rewritten from the ground up. From the hardware side, different kinds of memory, storage and sensor technologies will need to be created to serve neuromorphic devices. Both hardware and software will become completely different. Neuromorphic technology will imply the new digital revolution, with everything this term incorporates.
