AI With Purpose

Four Ways AI is Improving the World

Even before computers were invented, the idea of the “thinking machine” persisted. In fact, in 1872, the Victorian novelist Samuel Butler’s book Erewhon imagined a world in which machines would employ human-like intelligence to tackle society’s most pressing problems.

True artificial intelligence, however, has taken on different meanings and applications through the years. It gained momentum as a formal academic discipline in the 1950s, and while some strived to create an “electronic brain” capable of sentient thought, others sought to define the mechanical processes of higher logic so a machine could solve problems of unprecedented complexity.

In the 2010s, the advent of Deep Learning across neural networks ushered in a technical breakthrough for AI development. The driving force behind this AI revolution has been the democratization of the field through open-source collaborations as well as processors and servers with the computing power to unleash AI’s potential—both areas Intel has pioneered.

Here we take a closer look at four important areas where Intel’s AI technology is revealing its value to humankind and potentially revolutionizing industries.

1. Saving Lives with Early Cancer Detection

It cannot be stated often enough or loudly enough: Early detection of nearly every type of cancer can improve outcomes and save lives. For example, the U.S. Centers for Disease Control and Prevention (CDC) estimates that nearly 88% of adults who receive an early-stage diagnosis of colorectal cancer live for 5 years or more compared with 16% of those diagnosed at a late stage.

Cancer research is one of the most impactful use cases for artificial intelligence today, but like any other AI application, its success depends on the scale and depth of relevant data. A significant factor working against AI-driven cancer research at scale has been the confidential nature of patient data. However, Intel Labs and a leading medical university have devised a novel solution with the help of Federated Learning.

This international collaboration brought an unprecedented 71 institutions across six continents together, through a decentralized system using Intel Federated Learning and Software Guard Extensions. In a Federated Learning model, organizations can cross-reference updates from local data without sharing the data itself, meaning researchers gain the value of patients’ medical information without compromising their privacy. “By capturing learnings from data silos, Federated Learning can both improve existing use cases through increased data volume and diversity, as well as open up entirely new uses that were infeasible due to lack of training data,” says Intel Labs' Senior Researcher Micah Sheller.

The result: an incredible 33% improvement in brain tumor detection based on MRI scans from more than 6,300 patients. Other healthcare researchers can immediately benefit from this study through Intel’s open-source OpenFL project, which will allow anyone to deploy their own cross-silo Federated Learning applications with SGX. “I am honored to be working alongside medical AI researchers,” says Sheller. “Seeing our technology empower their efforts toward the common good excites me for the future of Federated Learning.”

2. Helping People with Visual Impairment See in Whole New Ways

The World Health Organization estimates that globally, 285 million people are visually impaired, and of that group, 39 million are completely blind. There is also a direct relationship between poverty and visual impairment in certain parts of the world. In fact, the NIH found that 90% of blind people live in developing countries in Asia and Africa.

Jagadish Mahendran, an engineer, developer, and startup founder originally used AI to teach robots to see. But a visually impaired friend inspired him to look at other applications. Jagadish was troubled by the high cost and limited utility of other offerings in assistive tech. So he combined satellite navigation, voice activation, and AI Computer Vision into a single wearable prototype that would lend a new set of eyes for the people who need them most. Using Intel’s OpenVINO platform to build and optimize his AI models for computer vision, Mahendran and his team are hard at work to bring an affordable build of the project to the market as soon as possible. Named Ximira (and previously known as just “Mira”), Mahendran’s prototype incorporates a backpack-housed computer and battery system, a vest-mounted spatial camera, and a pair of wireless headphones to provide up to eight hours of augmented navigation as the wearer traverses complex environments.

3. Restoring and Maximizing Crop Health for Vital Agriculture

In 2017, a combination of invasive microbial pests and plant disease wreaked havoc on farmers across India. Farms in the Orissa region experienced as much as a 90% crop loss. For Rishikesh Amit Nayak, an engineering student at a regional tech university, the problem was personal. His own family debated abandoning agriculture altogether as they struggled to chart a course ahead. With his classmate Niharika Haridas, he decided to use whatever data he could find to assess the problem with fresh eyes.

Using satellite-captured thermal scans of the area farms, they found some telltale signs of microbial infestations where the losses were most prevalent. With OpenVINO and a suite of AI development tools from their university, Nayak and Haridas were able to reliably identify early stages of plant disease so farmers could take immediate action before it spread.

4. Empowering People to Overcome Disabilities Like Never Before

For people with severe physical disabilities, things as basic as expression, communication, and access to information can be consistently out of reach. But for the past 26 years, Lama Nachman of Intel Labs has pioneered new tech in areas like Context-Aware Computing, Multi-Modal Interactions, and Human/AI systems. The result is Intel’s Assistive Context-Aware Toolkit (ACAT) an AI-powered system that allows people with severe disabilities to communicate through keyboard simulation, word prediction, speech synthesis, and more.

Nachman’s journey to developing ACAT began early in her tenure at Intel when her team was tasked with developing new assistive solutions for a longtime customer and collaborator who faced new health complications that limited his motor skills. And due to the nature of his condition, future-proofing this interface would require all-new methodologies and tech that simply hadn’t been invented.

“Somebody might be able to slightly move a finger,” Nachman says. “Somebody might be able to smile. [Is there] any possible way we can extract a signal through sensing and making sense of that data that can then map onto the equivalent of a push button?” By broadening the range of methodologies for “sensing” input and organizing them into an open-source system, Nachman’s team was able to offer a more sustainable long-term solution for this unique set of challenges. In the process, they paved the way for an open-source framework that could be configured for virtually any assistive use case.

Intel’s Vision for AI and the Road Ahead

Much of the technology being developed by Nachman and her team is being made available to developers around the world as open-source tools. “I have always been optimistic about the ability of technology to improve equity in the world,” Nachman says. “And the reason for that is that I feel like it can scale in ways that physical resources can't. When I think about the common thread for all these stories, I think we all see the power of AI, to be a transformative power for good.”

Written by Pikey Holdredge for Courageous Studios and CNN