AI and its pros and cons have been discussed to no end. Artificial Intelligence isn’t a new-age discovery. In India, the Centre for Artificial Intelligence and Robotics (a DRDO Organization) was established as early as 1986. So the question to ask is why all the fuss now?
When machines develop the ability to perform complex tasks such as audio/video recognition and decision-making that usually require human intelligence, they are said to possess Artificial Intelligence. Using robotics, these machines are even capable of implementing those decisions without requiring any human intervention.
Clubbed with new developments in the world of technology such as Big Data, the scope and future of AI has been altered and transformed significantly.
Businesses on a daily basis generate large volumes of data – both structured and unstructured. Earlier, most of this data went waste as we had no way of analysing it and working on it. With the advent of Big Data analytics, we are able to process and analyse these large data sets to discover meaningful patterns, trends, and associations that facilitate business decisions.
How Big Data impacts AI
There are 5 key aspects of computing that mark the rise of Big Data analytics in the technology world. These are the same reasons why Big Data is a critical enabler for AI implementation.
Exponential increase in computational power
Computer processors have seen exponential growth in computing speeds in recent years. Millions of data sets can be now processed in nanoseconds. In addition to sequential computing capabilities through CPUs (Central Processing Units), devices also have parallel computing GPUs (Graphics Processing Units). It is now possible to process large amounts of data in “real-time” and derive trends and rules for machine learning in AI applications.
Availability of low-cost and highly reliable large-scale memory devices
Efficient storage and retrieval of big data is now possible, using memory devices such as DRAMs (Dynamic RAM) and NANDs. Data doesn’t have to be centralised and stored within a single computer’s memory any more. Besides, we have too much data now to fit into one device anyway.
Cloud-based distributed data storage infrastructure allows parallel processing of big data. The results of these large-scale computations are used to build the AI knowledge space.
Machine learning from actual data sets, not just sample data
In the nascent years of AI, machines had to “learn” new behaviour from limited sample data sets, using a hypothesis-based approach of data analysis. But with Big Data, you don’t rely on samples anymore – you can use the actual data itself, available all the time.
Voice and image processing algorithms
Natural language processing, or understanding and learning from human communication, is a key requirement for true AI. But human voice data sets are voluminous, with scores of languages and dialects. Big data analysis enables breakdown of these data sets to identify words and phrases.
Similar is the case of image processing, which deals with recognising faces, shapes, maps, and boundaries. Big data analysis helps a machine recognise these images and shapes and learn to respond accordingly.
We have already begun to see this in action with the advent of Amazon Alexa, Apple HomePod, Google Home, and other virtual assistants.
Open-source programming languages and platforms
If you can store your entire data set in a single computer, then AI data models can use simple programming languages like Python or R, which are excellent for statistical data analysis.
But for commercial scale operations, companies might use big data management platforms such as Hadoop. It is a Java-based open-source software application framework which can read and analyse distributed data sets stored in clusters across different machines. The prevalence of reliable, free programming tools for data analysis have also made AI algorithm implementations easier and more effective.
Big Data/AI mergers – what’s next?
Recently, big data analysis at retail giant Walmart enabled them to take automatic business decisions. Walmart has about 245 million customers visiting 10,900 stores and 10 websites all over the world. The company collects 2.5 petabytes (1015) of unstructured data from one million customers – every hour. Using the data, Walmart analyses what customers are buying, what product is trending on Twitter, how weather might affect sales, and so on. Finally AI systems process the big data and make self-governing decisions such as:
- How many units of stock of each of their products are to be held in each of their stores
- Automatically place orders with suppliers depending in demand data
As this case-study shows, AI is all about analysing real-world data and helping the computer learn a thing or two from that data. When the learning is from samples or “small data”, it is the equivalent of clearing a snow-laden road using a tiny shovel – tedious and ineffective. However, when you use large real-time data sets or “big data”, it’s like a bulldozer ploughing through the snow in a jiffy – quick and immensely productive.
Artificial Intelligence and Big Data Analytics are two of the most promising technology paths that businesses can take in the future, to make intelligent decisions based on past business knowledge. But understanding the convergence and inter-dependence of these technologies is where the real success lies. We have only just begun to explore the possibilities of a big-data driven AI-powered world. Time will tell what these technologies hold for our future.