Per IBM, 90% of the data that we have in the world today has been generated in last 2 years!! Everyday we are generating 2.5 Quintilian Bytes ( 2,500,000 Terabytes) of data. This data comes in from all over the place such as social media, sensors, transactions, pictures, videos and so on. The growth of this data is expected to be even faster in coming decades. We need Machine Learning, Deep Learning and Artificial Intelligence to analyze this data to draw actionable insights and intelligence. This article gives you a view of what it means for us.
ML/AI is the subfield of computer science which tries to solve problems which are normally reserved for human (read Cognitive Computation) abilities. AI doesn’t mean that everything machines will be doing, rather AI can be better represented as “Augmented Intelligence”, i.e. Man+Machine to solve business problems better and faster. Even though most of us would have heard more about the ML/AI in last 4-5 years, this is not a new field. It has been flourishing from the late 80s / early 90s.
With the advent of Big Data/ Unstructured Data, faster computational speed, very cheap storage and need for customized and real time solutions, the ML/AI field is practically exploding now. Moreover, data is becoming Multi-Modal and Heterogenous. What this implies is that the data comes from different platforms and in all shapes and forms such as videos, text, images, social interactions, comments and so on.
We are already surrounded by apps and tools powered by AI which have become a common part of our daily life. We all are doing searches on Google every day, but may have not realized that the underlying algorithm for the searches is driven by ML/AI. Some of the other application of AI around us are Siri from Apple, Echo from Amazon, Cortana from Microsoft, Assistant from Google, IBM Watson etc
In not so very distant future, AI will continue to manifest itself in many more shapes and forms around us such as self-driving cars or generation of a movie trailer entirely driven by codes
We have heard about the Internet of Things ( IOT), but may not be aware that there is IIOT as well. This stands for “Industrial Internet of Things”. One key application of this is “Digital Twin”, which is continually learning digital replica of physical systems. It is estimated that connected machines and Digital Twins can save hundreds of billions of dollars via optimized operations
If you are job seeker with ML/AI skills, at least the next 10-15 years will be a booming period for you, where tech giants such as Google, Facebook, Microsoft, IBM, other companies across all verticals, and startups across the world will be equally interested in hiring ML/AI talent like yourself.
If you are an entrepreneur with a good ML/AI related idea, there will be plenty of opportunities for you to raise money to fuel your business growth. Per a Kalaari statistic, $6B USD has been raised by AI startups since 2014!
Classification and Categorization using statistical tools such as Support Vector Machine (SVM), Neural Networks, Naïve Bayes, Gradient Boosting Trees, Elastic Nets, K Nearest Neighbors (kNN) are used regularly for categorization of objects of interest.
Documents and news articles can be automatically classified into Politics, Sports, Entertainment etc. Images can be classified into animals, human, birds etc. Natural Language Processing- Sentiments (happy, sad, angry etc.) and context can be drawn from any text such as customer reviews and comments
Forecasting, regression, segmentation and other such statistical techniques are routinely used in ML for addressing different business requirements. Statistical Modeling- typically ML algorithms are built using supervised learnings, implying that we would need to build a model using training data (generally past data) and use these models to predict for future.
Programming- languages commonly used in the data science environments such as Python and R have a lot of in built libraries and packages for ML and normally used for developing the ML/AI algorithms.
Even with the advancements we have made in ML over the years, in certain cases there are problems which a 2nd grade student can solve faster than a machine. For example, identifying a person by looking at the face of the person. Any problems or questions which require social context will take longer for a machine to solve.
Particularly with respect to text analytics, there are two main challenges. First is “Ambiguity”. This means that the same word can mean many things. Second is “Variability”. Indicating the same thing can be said in many different ways.
Ontology and domain expertise is absolutely pivotal for any ML/AI algorithm to have high level of accuracy. Generally speaking, ML/AI algorithms developed for one industry may not be directly adaptable to another industry.
Democratization of ML/AI- Google, IBM and other such companies have made it easier for all of us to have access to and grow our knowledge on ML/AI tools and techniques. Some of the Free tools which you should try to take out for a spin are -
o Google machine learning stack www.cloud.google.com/ml and www.tensorflow.org
o Apache Spark www.spark.apache.org
o IBM Watson www.developer.ibm.com
o Microsoft Azure
In sum, the future of Machine Learning and Artificial Intelligence is really bright. We are just starting on a journey that may very well last for coming couple of decades.
Disclaimer: The views expressed here are solely those of the author in his private capacity.