Limited functionality available
Artificial intelligence is no longer a distant pipe-dream, it surrounds us every day. Some of us use cognitive technologies every day without even realising it. Siri, Google now and Cortana are part of many people’s daily livelihood. These popular applications are built off complex pipelines of information. To appreciate the black box behind these systems we must first understand the overall cognitive AI landscape.
At their heart, cognitive technologies aim to emulate human capabilities, providing a bridge between human consciousness and the static logic of computing. These technologies are growing to emulate every aspect of human thought processes, or direct emulation of our senses. Some categories such as Natural Language Processing (NLP) aim to emulate an entire system, from auditory and visual recognition through to complex language analysis.
Cognitive technologies have evolved out of the ever expanding artificial intelligence space. As AI become more common place within businesses, applications for the cognitive learning aspects become more apparent. Some of these system are consumer facing, such as Siri or Cortana, however many are not. Most cognitive technologies are shaped around streamlining operations within a business, creating value from information that was previously locked behind sheer volume or complexity.
There are many different Cognitive Technologies available already, including but not limited to; Natural Language Understanding, Machine Learning, Speech-to-Text, Biometrics and Handwriting Recognition. While these tools are still evolving, the information technology industry is yet to develop a strong demarcation amongst these diverse technologies. In order to categorise them, we need to understand their ultimate use within the business environment.
Uses for cognitive technologies can be broadly separated into three key pillars; engagement, insight and automation. From these key facets cognitive technology can be used to sense and shape processes, replicating and sometimes exceeding complex human thought patterns.
Many technologies within these categories can be adopted and utilised across almost any industry. When combined within a single business, these capabilities work together to enable integrated automation.
Cognitive technologies can be often applied in scenarios where the business engages with customers or end-users. Intelligent agents and avatars are used to amplify end-user experience by delivering mass consumer personalisation at scale, through communications methods natural to humans, such as visual and language.
Natural Language Understanding (NLU) technologies accept natural language from users and determine the intent of speech, allowing for downstream systems to catch and fulfill the user’s requests. On the other hand, Natural Language Generation (NLG) technologies convert structured data inside computer systems such as financial reports to a more human readable form, reducing the cognitive load for the user. Speech recognition and Speech Synthesis (Text-to-speech) technologies enhance this further with the ability to communicate verbally.
One example of a cognitive system improving engagement is RoboChat, a system by NAB owned UBank. This IBM Watson-powered Chatbot uses the Watson engine to help customers fill out complex home loan forms. The system is able to handle simple to moderately complex queries, allowing dialogue with the user through natural language such as “what is an interest-only repayment?” rather than a standard search engine query of “interest-only repayment’. RoboChat learns like any new employee; the system records issues and failures in customer service throughout the day. These issues that RoboChat “didn’t understand” are then analysed by human team members, who provide solutions directly to the program. This allows RoboChat to learn from its mistakes, increasing its knowledge base with every iteration.
Personalisation of user engagements is also a well-known application for many technologies in the cognitive domain. Personalised ads and direct marketing are often underpinned by technologies such as analytics and machine learning. Many organisations have adopted next-best-offer / next-best-conversation programs which use big data and machine learning capabilities to drive consumer behaviour based on their individual circumstances.
When combined with Artificial Intelligence (AI), Virtual Reality (VR) and the recently emergent Augmented Reality (AR) technologies offer the potential to delivery next generation customer experiences. The ability to navigate highly complex information in a personalised and simplified view, using visual, physical (such as hand gestures) and natural language mediums will provide a ubiquitous cognitive engagement amplifying human productivity and positive user experience.
These systems are capable of developing new patterns and relationships out of large scale data sources such as Big Data platforms. They are able to analyse in real-time, and create actionable, value-add insights from within billions of data points.
Cognitive insight technologies are capable of altering their algorithms and resultant outputs over multiple iterations (sometimes millions) without human input. Through these iterations the machine will alter its code, optimising the testing process for its next iteration. As this continues, the machine will retain successful processes, while culling failed processes. This results in recursive optimisation of the target system.
Netflix utilises machine learning to provide its every expanding user-base with curated recommendations far more complex than standard genre similarities. The system uses algorithms to interpret both the users’ history and general trends, sorting the user into a subset of “taste groups,” of which there are a couple of thousand sub-categories. These tastes are then matched against the ever expanding library of viewing options, and personalised categories and predictions are delivered to each user. This system is extremely successful, being responsible for over 80% of newly discovered shows an average user will watch.
Perhaps the most mature set of technologies, this encompasses those systems that are designed to automate repetitive, well-defined processes. These systems generally have a strong set of governing rules and libraries that create a rigid architecture around the “decisions” they make.
Cognitive automation is generally used to replicate simpler mental processes and activities. These processes are often rhythmic in nature such as content tagging, basic data extraction and rules based planning.
Intelligent automation is one of the most well adopted forms of cognitive automation. This focuses on the direct replacement of human processes via contained sets of computer software. Alone, IA is limited to parameters defined by their controller. However when combined with other techniques, such as machine learning, these processes may be maintained or even enhanced at a fully autonomous rate. This has been implemented through the Hong Kong subway, where an automated system plans and optimises over 2,600 maintenance jobs weekly for over 10,000 employees. This system calculates millions of different alternatives according to limitations such as train schedules and employee availability. It then determines the most optimal maintenance schedule each week.
Although we are in the infancy of cognitive technologies, it is clear that new capabilities will emerge and compound upon one another, as they did through the information communication boom. Cognitive systems work best when coupled with regularly updated source material. It is clear that the future of these systems lies coupled with other emergent technology such as Big Data and cloud computing solutions. Businesses able to utilise these systems in a cooperative space will gain the most value from investments into cognitive technology.
Harnessing the combined power cognitive computing and the evolving data management landscape may result in a wave of change across multiple industries. Many businesses will ride this change, adopting and embracing integrated automation as an extension to their current practices. Leveraging these systems will create an uplift company-wide, boosting efficiency and consistency of services and products created. This competitive edge will translate to direct value for customer, employees and ultimately company shareholders.
Marcus is an Analyst at Deloitte's Performance: Automate practice. He is experienced in Big Data solutions, DevOps and Agile software delivery. More recently, Marcus has been involved in the Automation practice, working on numerous viability assessments on workflow automation across Big Data sourcing, ingestion and analytic processes. This experience enables Marcus to deliver a unique understanding of the data requirements underpinning Robotic and Cognitive implementations within corporate environments.
Aung Htet is a Senior Consultant at Deloitte Performance Practice in Australia. He is an experienced data engineer and architect, delivering major projects in Financial Services, Energy and Telecommunication industries. With an early background in Database Development, Aung later focused on Big Data and Artificial Intelligence technologies. He is also a certified Robotic Process automation (RPA) developer and a member of the Deloitte RPA community of practice. Aung holds a B.Sc. (Honours) Computer Science from UNSW, with specialisations in Natural Language Processing and Robotics.