Cognitive Systems: Taking Cognizance of Machine Intelligence

Cognitive Systems

It’s been more than 60 years since the term ‘Artificial Intelligence’ or AI was coined. Since then it has provided great fodder to Hollywood movies like Ex Machina, The Terminator, I Robot, and many more, some portraying these self-learning machines as a scourge and others a savior. What powers AI is Cognitive Computing. In this blog, Rajeev Dadia talks about how Cognitive Computing has emerged as a disruptive technology fueling digital transformation.

The quantum of data being generated every second is unfathomable to the human mind. Every aspect of our life is being digitized with more and more attributes and frequency. Any consumer or buyer of goods, services and solutions is leaving behind a trail of data that depicts patterns and has insightful information. Deciphering intelligence from this data requires machines armed with specific techniques and algorithms.

What is powering these machines – Yes, AI.

But here we are talking about machines that could mimic the way human brain reasons, understands things in context and be complementing human abilities and even superseding in certain cases. Cognitive Computing has emerged as a disruptive technology where human and machines are collaborating to forge a new wave of understanding and thus building stronger and digitized enterprises. Cognitive Computing is already behind many of the technologies that we are using today – be it our internet based relevant search engines or personal assistants powering our smartphones – making our life easy and sorted.

Cognitive Computing can be simply understood as a software and hardware that learns “without reprogramming” and automates cognitive tasks. It is the combination of cognitive science (mainly Psychology and Neurology (and Computer Science and consists of Big Data and analytics, Machine Learning, IoT, NLP, casual induction, probabilistic reasoning and visualization). These systems not just make autonomous enterprise decisions but also  learn from their mistakes.

Cognitive Computing systems have supported breakthrough innovations in several industries. One such example is from the field of Oncology. Thanks to Cognitive Computing, an Oncologist can now scour through a world of data, and based on the tumor type, patients anatomy and demographic, determine the best evidence-based treatment within minutes.

A good candidate for applying Cognitive Computing is in the domain where a set of data (or a business questions on that data) may result in a hypothesis that has more than one possible answer. It is not necessary that answers have to be mutually exclusive. Same inputs may point out that a patient has two or more different issues or can develop multiple issues.

What makes it superior to its predecessors?

Here are three pointers that act as a ‘sort-of’ litmus test for a Cognitive system:
Contextual insight from the model, hypothesis generation (some sort of proposed explanation of a given observation) and last but not the least — continuous learning from data over time.

Cognitive system is modeled upon two of our core thinking systems — Automatic thinking which is based on intuition and biases and second that is controlled and rule-centric (also referred to as slow thinking)

Here is a high level view of elements in a Cognitive system (read each line as a layer):

  • Business Application – Cognitive system of engagement
  • Storyboard/ Visualization/ presentation layer
  • Model to Generate and Scope Hypothesis
  • Text Analytics
  • Descriptive, Predictive and Prescriptive analytics
  • Feature engineering, Deep learning, Natural Language Processing
  • Data Access, Metadata, Management Services
  • Internal and External data sources

Ultimately, the model which is at the core of this is based on “Continuous machine learning process” and modeled based on:

  • Repeat: Create a problem statement or take modified statement as input
  • Identify similarities from knowledge base
  • Generate Hypothesis
  • Identify Evidence
  • Score Evidence
  • Score Hypotheses
  • Present Result
  • Get expert input
  • Train and repeat

Although the core pieces of cognitive systems are part of our repository, the difference is between knowing the words and writing a poetry.
We need to get efficient at constructing sentences and then start working on bringing deep meaning and emotions in those sentences to make them poems. We need to simultaneously work on each line in capability and then take on low hanging problems and apply the whole continuous machine learning process and show “system is improving over time and with every run without re-coding” and claim use of cognitive system in full right.

Cognitive Computing is helping enterprises re-define themselves and transform into intelligent organizations. It’s not here to replace human intervention, but enables us to be better at whatever we are doing, thus, unleashing a new era of technology, thinking and intelligence.


About Rajeev Dadia

mmAs Saama Technologies‘ Chief Technology Officer, Rajeev Dadia drives the technology roadmap to enhance Saama’s technical edge and improve overall quality through leadership in delivery process and resource management. Rajeev focuses on cloud, syndicated data sources, social data, and mobile technologies as they impact the data and analytics needs of the organization. He has 20 years of successive experience in leadership roles at Saama and at Silicon Graphics (SGI) in Corporate Information Systems.

Related Posts

Rahul Sarma says:

Great article.
One of the best uses of cognitive analytics would be in finding relationships between various dataset that otherwise we may have ignored due to our traditional approach of only looking at attributes that are useful. Most of the times we filter out the attributes which are not useful for our datawarehouse and hence limiting our ability to correlate. Its possible to do manually but again is a very painstaking process. A cognitive platform would automatically help connect dots which we did not know existed.

Leave a Reply

Your email address will not be published. Required fields are marked *