Deep Learning: Evolving the 'intelligent' machine

Every day, 182.9 billion emails are sent and received worldwide. Did you know that 65% of all email is spam? With help from advanced algorithms, computers continuously recognize and segregate this mail so it does not feature in our inboxes. It is this ability to ‘learn’ from large blocks of data – and evolve classification methods accordingly—that lies at the core of deep learning.

This valuable technology, whose applications (such as spam filters) often feature in our everyday lives, has come a long way from the days when Supercomputer Watson defeated the world’s strongest Jeopardy champions.

Deep learning uses artificial neural networks or computational models that resemble the intricate neurons of the human brain. Together these networks ‘learn’ from data and recognize patterns that ordinary computing would not be able to do, for example, in areas such as speech recognition and computer vision. Today, it takes on important classification, prediction and decision-making roles that experts say is furthering success in creating a fully functional Artificial Intelligence (AI) system.

The healthcare industry, faced with an overwhelming number of regulatory requirements and the need to maintain accurate and up-to-date patient records has turned to deep learning to provide tech- based assistance. The technology can be implemented to identify individuals who display a higher propensity for certain illnesses by programs that study the vast base of electronic health records. Based on the results pro-active action can be taken particularly when lifestyles are successfully matched with medical conditions.

Pattern recognition can also play a crucial role in the law enforcement sector. MIT scientists have developed algorithms that tackle the labor-intensive task of comparing criminal history and area statistics with recent cases. Data-driven techniques automate this process and can help the police apprehend suspects with increased speed and accuracy.

The benefits of deep learning are also becoming more accessible to individual consumers and small businesses through open source, customizable software. One platform allows users to build smart applications that pre-empt user behavior based on data science. Machine learning algorithms that are scalable and utilize existing big data networks can create predictive features such as personalization, recommendations, and content discovery for a number of digital properties.

Researchers are particularly excited about deep learning and the potential it holds in establishing a truly AI system. With milestones such as Project Adam becoming the world’s best photograph classifier the potential disruptiveness of this technology is immense. Soon cellphone cameras could play multifaceted roles ranging from helping blind people ‘see’ to providing users with nutritional information of a meal from just its photograph.

Do you think deep learning has the potential to usher in the next wave of intelligent machines? Leave your thoughts in the comments below.

About the Author

Wipro Insights set up the Council for Industry Research, comprising of domain and technology experts from the organization, to address the needs of customers. It specifically looks at innovative strategies that will help them gain competitive advantage in the market. The Council in collaboration with leading academic institutions and industry bodies studies market trends to equip organizations with insights that facilitate their IT and business strategies. http://www.wipro.com/insights/

Email us at: wipro.insights@wipro.com

Service Offerings

Client Themes

Service Offerings

Client Themes

Our Global Presence Global Site

Our Global Presence Global Site

Our Showcase Pages Locations

Deep Learning: Evolving

the 'intelligent' machine

About the Author

Service Offerings

Industries

Digital

HOLMES

Innovation

Contact Us

About Wipro

Location

Leadership

Investors

News and Events

Blogs

Partner Ecosystem

Careers

Sustainability

Contact Wipro