Blogs

What is cognitive IoT?

Post Comment
CTO, Internet of Things, IBM

Put simply, cognitive IoT is the use of cognitive computing technologies in combination with data generated by connected devices and the actions those devices can perform. You probably already know what the Internet of Things is about and what we mean by sensors and actuators. In focusing on the cognitive computing aspect, what does it mean for the Internet of Things? Cognition of course means thinking, and while computers are not yet capable of general human-like thought, they can now perform some of the same underlying functions that humans perceive as thinking. Cognition involves three key elements: 

  • Understanding
  • Reasoning
  • Learning

In a computer, system understanding means being able to take in large volumes of both structured and unstructured data and derive meaning from it—that is, establish a model of concepts, entities and relationships. Reasoning means using that model to be able to derive answers or solve related problems without having the answers and solutions specifically programmed. And learning means being able to automatically infer new knowledge from data, which is a key component in understanding at scale.

http://www.ibmbigdatahub.com/sites/default/files/cognitive-iot-blog.jpgBuilding complex models of concepts and relationships at scale can be too time-consuming and costly. Furthermore, many relationships are not known or obvious beforehand, so they are only practically discoverable by having a machine automatically analyze large data sets to discover patterns.

Things that think

Cognitive computing is significant to the Internet of Things for a few critical reasons.

  • Rate and scale of data generation: Learning helps optimize processes or systems to make them more efficient based on combining sensor data about the system with other contextual information. The data generated from devices can quickly overwhelm the human ability to analyze for detecting important patterns and learning. Applying machine learning is essential to being able to scale the Internet of Things.
  • Computing’s movement into the physical world: As more people of all ages and technical skill levels interact with Internet of Things systems, we need to move beyond current machine interface paradigms that require humans to learn the abstractions and specialized interfaces needed to interact with machines. And that movement needs to be toward a more human-centric interface. In other words, people need to be able to interact with Internet of Things systems—things—using natural language. The systems have to begin to understand people. Author David Rose from the MIT Media Lab coined the term “enchanted objects” to characterize the seemingly intelligent behavior that we can infuse into connected devices through the Internet of Things and cognitive computing.
  • Integration of multiple data sources and types: In the Internet of Things, many data sources exist that may provide related information or context for better understanding and decision making. The ability to digest and analyze different types of data, including digital sensor data, audio, video, unstructured textual data, location data and so on, and to identify correlations and patterns across these data types are very powerful capabilities. Understanding the intention of human operators can be greatly enhanced by greater knowledge of the context—physical context, temporal context and even emotional context. Reasoning and decision making can be improved by integrating multiple different data sources—for example, correlating sensor data with acoustic data.

Human-aware devices

Cognitive IoT is the next leap in improving the accuracy and efficiency of complex, sensor-driven systems through learning and infusing more human awareness into the devices and environments we interact with. This leap can make our things understand and interact with us in our language(s) instead of the other way around.