13 Oct DEEP LEARNING
What is it?
Deep Learning is machine learning using a set of algorithms that can allow a computational model to infer meaning from fed information, often speech, images or text. It’s a deeper form of pattern recognition than what is being used commercially in the world.
Machine learning is shifting from a highly manual process where humans have had to design representations for each task of interest, to an automated process. In this automated process machines learn more like kids do, through experience, making sense of the data they encounter by automatically building logic based representations and using these simple representations to then build more complicated ones.
Why is this trend happening?
A lot of corporate value in the world is contained in companies that use algorithms and pattern recognition in their products and services. Google which made a splash with its half a billion dollar + acquisition of Deep Learning startup Deep Mind is the second most valuable company in the world getting majority of its revenue from search based advertising that uses such techniques. Apple’s Siri and IBM’s multi billion dollar investment in Watson are dramatic examples of where Deep Learning capabilities are being used and bet on. Deep Learning promises to revolutionize fields and capabilities across industries because at its heart is the promise of smarter analysis.
What are the benefits?
Deep Learning helps you understand complex data sets. The automation of the machine learning process coupled with access to the amount of data generated powered by ever increasing advances in computing power will enable Deep Learning to benefit every industry to do its core tasks better.
Investors are already using Deep Learning techniques to get better returns on Hedge Fund investments. Search via speech or text is a critical component of easing the burdens of modern life and eliminating wastages from it. Deep Learning enables that search. It has great security implications too as Law enforcement Facial recognition programs all rely on it.
In the future Deep Learning could make possible instant translation in real time bringing the entire world closer. Deep Learning will make self driving cars possible and allow for incredible efficiencies in all aspects of daily life for each person.
What are the challenges?
To go beyond text, speech and visual data related applications the amount of computing power required is even more then current standards. Finding qualified people to advance the cause of Deep Learning is also notoriously difficult. Simple data scientists are in short supply and a world class Deep Learning expert that often command million dollar salaries are proportionately more rare. To get this kind of talent on board and give it access to relevant computing power is the purview of a handful of companies in the world. Deep Learning field thus has a chance to be hijacked to fulfilling the commercial goals of these companies. Ground breaking advances in the field will come when research into its critical sub parts like “high dimensionality, streaming data analysis, scalability of Deep Learning models, improved formulation of data abstractions, distributed computing, semantic indexing, data tagging, information retrieval, criteria for extracting good data representations, and domain adaptation” is sponsored. But commercial inclinations could delay that.