Thinking Machines: Has the Era of Cognitive Computing Arrived?

Blog /Thinking-Machines-Has--the-Era-of-Cognitive-Computing-Arrived

I recently had an interesting conversation with a chief architect about the hype around cognitive computing and its relevance to data monetization efforts for analytics. The chief architect said he believes data cannot be monetized without cognitive computing—an interesting theory, given that monetization of data is impacting all kinds of businesses today.

But I began to think about how cognitive computing might change the data analytics space, which is already overwhelmed with big data.

What is cognitive computing?

Cognitive computing technology is not exactly a new concept. If you, like me, studied artificial intelligence (AI) in college, you understand the basics of cognitive computing. AI was a favorite discussion topic in my college days, despite the fact that we did not know how to make it a reality.

In those days, we heard that computer scientists were building systems that were as intelligent as humans. One element of these systems was natural language processing (NLP), which introduced machine-learning algorithms for language processing.

Today, we have modern and common machine-learning algorithms like linear regression, logistic regression, k-means, naïve bayes, random forest, K-nearest neighbors, support vector machine, decision tree, and dimensionality reduction to support the data analytics space. And a couple of years ago, data analytics  evolved further with the advent of data lakes—a way to store huge amounts of raw structured, unstructured, and semi-structured data in native format.  

So, in essence, cognitive computing is a blend of AI, NLP, machine-learning algorithms, ontology, data ingestion and analytics involving data lakes. To a data management practitioner, cognitive computing is about getting insights from data management platforms and data-centric applications and blending them with human intelligence; data management best practices for structured, semi-structured and unstructured data; analytics; and context-based analytics.

With today’s powerful processors and storage capacity, cognitive computing may be poised to revolutionize data analytics.

What’s driving cognitive computing applications?

The first driver is the volume and growth of web pages and applications. Data from intelligence devices is also a major driver. The third driver is cognitive computing’s appeal for analytics. (After all, data is growing more complex by the hour.)

The fourth driver is industries like healthcare and life sciences. Experts in these fields are pushing for cognitive computing solutions that can help them with complex medical diagnoses. Rather than trying to keep up with all the published articles and research studies, they want to be able to quickly make decisions in few hours (instead of months) using data analytics solutions built on relationships and connections across tons of articles, journals, and research papers.

What are the biggest challenges to developing cognitive computing applications?

It goes without saying that will be something of a learning curve for each element (AI, NLP, machine-learning algorithms) of cognitive computing. But the real challenge for cognitive computing is our ability to train applications or systems to look at unclear data elements and their relationships.

What do cognitive computing applications need to be successful?

As with any data analytics project, the reliability and quality of the data is key to success. Analysts talk a great deal about big data, but what’s really needed is relevant data to analyze for pattern and analytics. It is risky to mix all kinds of data because it may end up not providing any meaningful insights.

There is also a tendency to jump into big data and data lakes, but it’s best to start by defining a domain-specific problem, formulating a hypothesis, and ingesting small set of data—a small bucket of data rather than a data lake.

Finally, it’s important to understand the connections between data elements of data centric applications, along with volume and variety of data. For example, there are critical connections established between data elements of graph data used in social media sites like Linkedin and Facebook.

Of course, while domain is important, we must first understand the problem we are trying to solve.

It’s still very early days for cognitive computing. As we consider projects involving it, we must remember the three elements needed to succeed with any technology:

  • Specific domain
  • Specific focus
  • Specific goals

Unless we know the problem we are trying to solve, the best technology in the world won’t get us to a solution.

Fecha de publicación: 24/08/2016