Big Data is already valuable but, according to a report on IBM’s Big Data & Analytics Hub, it could become even more so if deep learning algorithms live up to their promises. Here, RT Insights founder and Executive Editor Les Yeamans explains how this technology can give computers the ability to recognize and respond to sensory patterns in a way that comes naturally to all sentient beings.
Already, these algorithms are learning how to recognize patterns in video, audio, speech, image and other non-textual data. This is reflected in things such as voice and face recognition software, and the technology will soon become commonplace in our smartphones, wearables and in the Internet of Things (IoT).
Streaming media is the focus of deep learning algorithms and they are getting better and better at recognizing faces, objects, gestures and voices – but that’s only scratching the surface. To be truly valuable, deep learning needs to be able to not only identify what is in the picture but to understand what is going on as well. Can an algorithm possess true insight into what it’s analyzing?
Deep learning that can generate that insight by corroborating with additional contextual variables will be the game changers. These variables would be found by looking at the attributes of the people and objects in the stream or image being analyzed and factors, including these categories:
- Social
- Geospatial
- Temporal
- Intentional
- Transactional
By working together, a person could be identified as being at a specific place at a specific time. Researchers would like to reduce the need for deep learning to rely on extrinsic content for full analysis but that is not something that is likely to happen anytime soon. It’s thought learning to leverage the data from other types of machine learning algorithms such as natural language processing and behavioral analytics may put them on the right path.
The situational variables identified by deep learning could be woven into narratives to describe the journey (literal or symbolic) that the individuals in the analyzed stream may be on. Some are making claims that they can identify abstract concepts such as “fun” but skepticism remains. That might be asking too much at this point in time.
There will always be a need for human judgment of these auto-generated narratives to ensure they ring true with both the viewer and the people involved in them. If the narrative and human disagree, the human judgment trumps all.
