READY FOR THE NEXT CHAPTER IN AI?

Image recognition is currently used in quality control (to detect defects), security (scanning faces and car license plates), and medicine (tumor detection).

One problem organizations often face is providing enough data to train image classification and recognition algorithms, and pre-processing images can take more than half the total time to resolve. Intel® Xeon® processors can support data augmentation applications to solve this problem. Such applications rotate and scale images and change colors so that fewer images are needed to effectively train image recognition algorithms (depending on the use case).

Thanks to their energy efficiency and high memory bandwidth, the CPUs easily handle data augmentation applications. This is especially true for the scalable Intel® Xeon® processor series, where processing is accelerated by Intel® Advanced Vector Extensions 512 (Intel® AVX-512) instructions.

Voice-activated virtual assistants can not only fulfill requests accurately, but also "understand" the nature of the requests to continuously improve. Similarly, customer satisfaction can be enhanced using systems that can process call center records or handwritten forms - a treasure trove of previously hidden insights that can be used to identify common complaints or respond more quickly to customer issues.

Natural Language Processing (NLP) uses the technology known as "recurrent neural network" (RNN) and "long short-term memory" (LSTM), and Intel® Advanced Vector Extensions 512 (Intel® AVX-512) comes into its own when processing the loops and dependencies that characterize these operations.

• What is the appropriate use case for my organization/company?
• What information technology do I need and what is already available?
• Do we have the necessary know-how, and if not: where can we find these competencies or how can we develop them?

Predictive maintenance differs from image recognition and natural language processing techniques in that it typically operates at much lower data rates and data is collected from sensors to monitor endpoint operating conditions. Ideally, as much of the processing as possible should occur at the end-device level before data is transferred to the cloud for analysis or decision-making. The Intel® Movidius™ Neural Compute Stick equipped with a VPU (Vision Processing Unit) is ideal for deep learning acceleration at the edge.

and if it makes sense to expand existing capabilities, using open source frameworks with your existing data center architecture can greatly simplify AI adoption across the enterprise.

aI performance is determined by a combination of compute power, software optimizations, and processor memory bandwidth; and regardless of how far along you are with your AI implementation, Intel's portfolio of hardware and software provides you with a rich set of tools for building the most cost-effective architecture for AI tasks.