IBM Injects Fresh Dose of Deep Learning in its Cognitive Computing Platform, Watson

By CIOReview | Tuesday, July 21, 2015

FREMONT, CA: Technology developments are assisting breakthrough innovations in Artificial Intelligence (AI). Computers have turned smarter, but not truly intelligent. In the quest of guiding artificial intelligence in a promising direction, IBM has added a powerful AI technique, known as deep learning, to its commercial version of Watson, reports Will Knight for MIT Technology Review.

Watson is an artificially intelligent computer system from IBM, capable of replying question posed in natural language. The recent acquisition of Alchemy API has enabled IBM to empower its AI with deep learning. Deep learning involves training a computer to recognize complex and abstract patterns by feeding large amounts of data through successive networks of artificial neurons, and refining the way those networks respond to the input, reports Will Knight for MIT Technology Review.

The company is integrating three deep-learning based features to the Watson API such as translation, speech-to-text and text-to-speech enabling apps and websites to build on to offer translation and transcription services.

Online availability of classified data has enabled advancements in deep learning and powerful graphics processors have aided their performance in crucial computations. Though deep learning produces striking results, lack of real world perception can lead to failure. Effective integration of deep learning with other AI techniques can produce better results.  Jerome Pesenti, VP of core technology for Watson imagines an intelligence which redirects to actual human when the system falls short of answering a query.