Realize business value from big data with Infosys data analytics solutions.

Results tagged “AI”

AI Reborn

 

During my academic days we had a dedicated subject for Artificial Intelligence (Al) in our Final year. Of what I could remember, it was all about Algorithms and a unique less known language called LISP (list Processing) for which we had labs back then. It was one of my favorite subject as it was futuristic and not normal science. But coming out of college I believed that Al was something that would continue to be a research topic rather being a practical implementation. Why? Because how would you come up with an algorithm that could mimic Intelligence? It is one thing to write a code for Chess that could analyze all the "n" possible moves and choose the one that has the smallest high probable path for victory but it's totally a different case to provide a generic Intelligence. Those were the days post IBM Deep Blue's victory stories against Human Grand Masters. So, called intelligence in domain like chess is quite possible as it has boundaries. There are limited rules and probable moves. A good set of algorithm and powerful processing power would give any grandmaster a run for his money. But in a more variable domain, things get more fiction than being realistic.

Fast forward seventeen years and Al seems to be more realistic and evolving towards realization. And I get a feeling YES, it's possible. So, what has changed? To answer this in two words -"Data Analytics".  Today's AI is driven by Data Analytics, Algorithms being developed are focused more on Data. Other than this most of the advancement in Today's Information Technology landscape is aiding Data Analytics.

The foundation of Today's AI has been data; algorithms under the field of Machine Learning and Data Mining are focused around data and harvesting patterns from it, these harvested patterns forms the building blocks of today's AI. Unlike yesteryears, today data is available in huge volume, the digital world around us has been spewing data all around, and this data in form of Big Data provide the mining field for patterns which usually remain unseen on surface until right analytics has been applied. With real time analytics, intelligence no longer need to be harvested from historical data but can be attained and recognized as it happens.

Transformation in technology landscape has aided the evolution of AI as well. Cloud computing is providing virtually unlimited storage and computing power required to run data analytics at scale. Another crucial element that is contributing to the success of AI today is "Connected System" through Web services and API's. AI yesteryears was conceived to be more of a monolithic, an AI system back then was supposed to house most of its intelligence capability, but today intelligence is distributed. Today if you want your system to recognize human speech than you need not build it from scratch but can leverage existing speech recognition services from Microsoft, Google or Amazon. Similarly if you need an image recognition capability than you can look forward for Vision API's again from players like Microsoft, Google and Amazon.

Just like distributed architecture the evolvement of AI has also been distributed and community driven. Popular statistical computing language "R" widely used for Data Analytics is open source. What makes "R" so popular for data analysis is its vast range of packages developed and contributed by independent developers solving certain problem domains and making it available for reuse this collective effort has been contributing towards the advancement of Data Analysis and AI.   

These are some of the factors providing the right ecosystems that is making AI to thrive and become a reality today. As hinted above no one entity is building the whole AI ecosystem but it's been evolving gradually in bits and pieces. Bunch of these services are baked locally to provide certain AI implementation like "Self Driving Car". Gradually such focused implementation can be augmented with one another to give rise to more intelligent system with broader capability that could someday rival humans. And probably one day, tables might turn around and machines might be working on making humans more intelligent and efficient so that we humans can serve their purpose, Scary uh!

Deciphering the Minority Report on AI

If Thomas Friedman decides to revise his best seller (The World is Flat) in the future, I think he would include AI/ML as an important disruptor. Per Jeff Bezos, Artificial Intelligence is in its golden age. Jeff calls AI an enabling layer that will improve every business. At the World Economic Forum in Davos, Satya Nadella said AI could be a vital driver for growth. Mark Zuckerberg predicts AI to deliver many improvements in our lives in the next 10 years. Google co-founder Sergey Brin says he's 'surprised' by pace of AI and calls it revolutionary.

But not everyone seems to be on the same page. Not Elon Musk for sure. Last year, he compared AI to summoning the demon. This year he went ahead and called it the biggest risk we face as a civilization. While the majority seem to feel positive about AI, it's hard to ignore the minority report, especially when it comes from one of the most respected visionaries in the current times.

So, what is the truth? Is AI really a threat to our existence? Categorizing machines into 4 kinds and evaluating the risks introduced by each of these, I have tried to find an answer to the earlier question.

  1. The flawless clerk
  2. The expert system
  3. The invisible machine
  4. The silicon poet

For the complete perspective, please click here

As our CEO, Dr. Sikka says AI - Pursuit of building something intelligent is as old as humanity. It has grown leaps and bounds from the time AI was discussed in 1956.

Since its inception for what determined a machine to be "intelligent" AI has evolved overcoming challenges during 90's and has entered its golden era.

Increasing investments from Nvidia on GPU and Google in TPU has made Moore's law more and more relevant in current scenario. Availability of these powerful processing units have accelerated Deep learning, Automate Machine learning and artificial neural networks.

Startups have revolutionized AI world and are providing disruptive solutions to solve complex business problems, these days every organization is adapting AI in some or other forms, may be as simple as for customer segmentation, social integration, personalized offers, supply change or building complex solutions for solving human problems around cancer treatment, self-driving cars and many more.

Over next couple of years maturity of AI will rapidly increase and more unconventional use cases will turn into reality for consumption.

Here is first in series of few such use cases in my view as Infrastructure, platform and products becomes available to implement.

Emotional AI:

Deep learning is the study of artificial neural networks related to machine learning algorithm containing more than one hidden layer similar to human brain. Based on deep learning AI can distinguish between dogs and cats, good vs criminals but how do you feel when you are treated from a robot, how does families feel when AI is able to identify depression in struggling students?

Among lot of other things AI is able to recognize faces, turn sketches to pictures, identify voice and many more.  

Days are not far when Siri or Alexa can detect emotions based on the sound of your voice and have a conversation, recommend a therapy session or send an alert to your loved ones to order a bunch of flowers to make you happy.

As human beings, we understand contexts and empathy. Not many AI models have it today. Companies that can implement these into their technology will have more success.

AR Chatbot:

2017-05-29-PHOTO-00000140.jpg

Chatbot is common platform these days and everybody has a version of right from Microsoft, Facebook, Watson, Google or custom built.

People are probably going to be more drawn into engaging with chatbots which has personality; it has to be companion to whom people can engage with.

AI integrated with augmented reality can do wonders. If bots can be integrated with AI, emotional AI and AR then humans-robot interactions will take a huge leap forward. Humans often struggle with appropriate responses due to complexity of emotions, if technologies can decipher this then the output will be very impressive.

To be continued...
1