The promise of AI is driven by memory and fast storage

The discussion of AI is everywhere. It was a key theme of seemingly every tech conference last year, and nearly every company, tech and non-tech, is talking either about where they play in AI or how they will take advantage of it to improve their business. The reason is clear: the promise of AI is real, and the potential is bigger than most people realize.

Hundreds of new solutions are already underway that illustrate the pervasive and dramatic impact AI will have. Let me give you a few examples.

• Today, there are new smartphones that use a simple AI to recognize a scene and adjust camera settings to capture the perfect picture. Taking better smartphone photos may not be life-changing, but it’s a good example of the subtle ways AI will quickly become a natural part of our everyday life. In fact, you likely already interact with AI results every week through the recommendation engines that drive streaming video, music, and online shopping systems.
• One of the other most prevalent uses of AI today—speech-recognition—is expected to multiply dramatically in the years ahead. In just five years, some reports project we will have 10 times as many voice controlled devices as we have today.
• Researchers have proven that cancers can be fought more effectively through DNA sequencing and analysis by powerful computers. The Cancer Genome Atlas is building the beginnings of an index that will help identify personalized and much more effective cancer treatments.

The drive to implement these new solutions is already changing Micron’s business. The memory and fast-storage solutions we make have always been central to the basic functions of computing, but they have not always been viewed as central to the value that computing created. AI is helping change that perception. Now that data and data analysis are the measures of value creation and differentiation, a computer’s ability to quickly move, interpret, and make sense of that data becomes paramount. Consider the records Nvidia’s DGX-2 systems set this December in MLPerf, the industry’s first broad AI training benchmark. The benchmarks cover five categories of important AI training tasks (image classification, object detection, translation, recommendation, and reinforcement learning). The DGX-2 systems set six new performance records, and they relied on massive capacities of high performance memory in each system (1.5TB of DDR4 and 512GB of HBM2) to do so.

The increased relevance of memory and fast storage to these essential AI tasks is giving Micron’s researchers and engineering teams new opportunities to achieve our vision — to transform the way the world uses information to enrich life. An exhilarating future is ahead of us; I cannot wait to see what’s next.