In the course of my work as Intel’s director of machine learning strategy and outreach, I’m often engaged with customers who want to put artificial intelligence to work to solve hard problems. This was the case recently when a healthcare provider turned to Intel for help in developing an AI system to augment the work done by its doctors.
In this particular case, the healthcare provider wanted to improve the accuracy of doctors’ diagnoses of a potentially deadly illness that can be hard to diagnose because of common symptoms that can be caused by many different conditions.
Here’s where AI enters the healthcare picture. To speed diagnosis of this illness in patients, the healthcare provider wanted to create an AI system that could quickly analyze huge amounts of data and make a fairly accurate prediction. People have wanted to use AI to do things like this for years, but until recently the technology wasn’t there—at least not in a system that would be both accessible and affordable to institutions like hospitals and emergency care facilities.
Today, that technology is all around us in the form of distributed x86 computing systems that can scale out seamlessly to allow many processing nodes to work on a problem at the same time. In this case, the healthcare provider considered using a GPU-based system to tackle the problem, but then determined that such a system could not accommodate the humongous amount of data it would need to train models on short timelines—because that data would all have to fit into the memory in a single node.
This is when the customer turned to Intel for help in developing a scale-out AI system based on the Intel® Xeon Phi™ processor family. This distributed memory architecture allows AI models to run over many nodes to accelerate the training of AI algorithms. We think that a distributed processing approach on industry-standard servers is clearly the future for AI. It puts AI into the reach of many more organizations by enabling them to leverage their existing x86-based infrastructure to power machine learning, rather than forcing them to make significant hardware investments to access AI capabilities
While I have used the healthcare industry as an example, the push to adopt AI systems is one that spans virtually all industries—from transportation and manufacturing to finance and entertainment. In the near future, many organizations will leverage machines that sense, learn, reason, act, and adapt to the real world. These machines will amplify human capabilities, automate tedious and dangerous tasks, and solve pressing societal problems.
Here are a few examples of the way AI is transforming industries:
|·Advanced drive assistance
· Personalize experience
|· Performance optimization
· Injury prevention
· Fitness management
|· Smart factories
· Smart cities
· Precision agriculture
· Power management
|· Early diagnostics
· Precision medicine
· Drug discovery
|· Fraud detection
· Risk assessment
· Asset management
At Intel, we are excited about the potential of AI to transform our world in amazing new ways. To that end, we are working actively with a broad ecosystem to accelerate the development of new technologies and solutions that will make game-changing AI capabilities accessible to a broad range of institutions. Our vision is the democratization of AI—a shift that will transform businesses and change our society in fundamental ways.
At the Intel Developer Forum this week in San Francisco, we’ll be talking with industry thought leaders from Google, Facebook, Microsoft, and more about the trends, challenges, and opportunities around AI and advanced analytics. We’ll also be providing technical insights on how organizations can scale their data centers to meet the growing needs of AI. For a closer look at the work Intel is doing to accelerate the development and adoption of AI systems, visit our machine learning site.