One of the most exciting (and for that matter promising) disciplines in the field of artificial intelligence (AI) is deep learning. Deep learning can be defined as a new area of machine learning working to improve areas like voice search, image and language processing or solve unstructured data challenges. You may not be aware of deep learning, but it is everywhere around us - think of the personalized recommendations you get while shopping on Amazon or using Siri (or another voice-activated assistant).
Here are three of what I think are the most crucial areas of deep learning:
Voice Search & Voice-Activated Assistants
Voice-activated assistants can now be found on virtually every smartphone thanks to the big investment the tech giants have made in deep learning. Apple's Siri was introduced to the market October 2011. Google Now, the voice-activated assistant for Android, was introduced a year later. The goal is to quicken the actions you already do on your phone and other devices by learning from your behavior. Google Home and Amazon Echo are voice-activated household devices allow you to play music, calculate a math problem, even turning off lights, to name a few.
This article from Fast Company reviews some of the voice-activated contenders.
Image Recognition
Image recognition, or computer vision, aims to recognize and identify people and objects in images, as well as understand the content and context. There are programs that help the visually impaired, safety features in cars that detect large animals, auto-organize untagged photo collections and extract business insights from socially shared pictures.
Using deep learning techniques, computers can be taught to accurately identify what is in pictures faster than ever, but they need massive amounts of data to do it.
Recommendation Engines
The two main industries benefiting strongly from recommendation engines are the retail and media industry because both have a lot of data in the long tail and can overcome the cold-start problem. You may be scratching your head wondering what long tail and cold-start mean.
Long tail: Traditional retail economics only stock likely hits because shelf space is expensive. Online retailers (from Amazon to iTunes) can stock virtually everything, and the number of available niche products outnumber the hits by several orders of magnitude. Those millions of niches are the Long Tail, which had been largely neglected until recently in favor of the likely hits. Amazon is a good example of a business that does amazingly well from the long tail.
Cold-start: At the heart of a recommendation system is that a computer learns from data, i.e., who has viewed this movie before (Netflix), who connected to this person before (LinkedIn), etc? One of the challenges can be that there is not enough historical data from the start, thus the algorithm will not be able to recommend anything useful in the beginning as there is not a long purchase history.
Deep learning is becoming an important area for all types of economy sectors. Some impressive applications include:
- Navigation of self-driving cars
- Predicting the outcome of legal proceedings
- Precision medicine
Originally published on LinkedIn.