Page 36 - Cyber Defense eMagazine December 2022 Edition

P. 36

driver to take a “short cut” down a side street, Google is creating a dataset of training data to enable the
development of autonomous vehicles.

The development of AI is patently different from the development of any other modern technology. Where
most modern technologies rely on basic if/then decision trees, AI creates a neural network of innumerable
datapoints to mathematically calculate the best solution to any problem. For this neural network to
function, the AI must be “taught” to make connections across a vast universe of data. And unlike many
technologies, the success of a particular AI will be determined by the quality and quantity of data it can
access.

You are already seeing AI at play when you turn on Amazon Prime and have custom video content
queued up based upon the genre of your recent book purchases, or the political videos appearing in your
Facebook feed after you view the election results on MSNBC. Soon, this type of data-driven personalized
experience could apply to every part of your life.

Take healthcare, for instance—

With access to quality data, in the not-too-distant future, health monitoring AI would be able to identify
and diagnose the onset of illness and disease in ways modern medicine is simply not capable of.
Smartwatches and mobile devices will be able to work in concert to identify imperceptible symptoms such
as irregular breathing, sleep patterns, heartrate, and a change of gait. These symptoms would be flagged
as anomalous against a health data baseline created from years of 24/7 monitoring from your devices. It
would also be checked against the baselines of millions of other people of a similar age and demographic
worldwide. The AI at the backend of this platform would incorporate datapoints from your genetic code,
as well as your medical history and the medical histories of your immediate and extended family. Using
location data from a mesh network of mobile devices, the AI could also determine who in your recent
proximity might have exhibited similar symptoms. Based on near-instantaneous analysis of your personal
data as well as a thorough understanding of the entire compendium of medical studies and research, the
AI could diagnose ailments at the earliest possible moment and request that your doctor approve a
recommended prescription available for immediate delivery to your location.

You wouldn’t necessarily know why the AI is making the decisions it is making, but it would create a
decision web from millions of datapoints to achieve the best outcome for your wellbeing. Expanding this
web to the Internet-of-Things, your home and office thermostats could lower the ambient temperature to
account for the coming fever; your office calendar could automatically reschedule the next morning’s
meetings; and your refrigerator could order Pedialyte, Tylenol, and chicken soup. Before you know you’re
sick, you could already be on the path to recovery.

A future like this is predicated on technology developers being able to access immense quantities of both
high-quality training data for AI development and the sharing of data collected by multiple sources to
create the necessary digital neural networks. However, in the United States and other Western
democracies, much of the data required to achieve the level of personal automation in the above scenario
is currently neither centralized nor shared freely across organizations, who prize this data for its
commercial value. Moreover, some of the data is also governed by regulations—such as the Health
Insurance Portability and Accountability Act (HIPAA), CCPA, and GDPR—which prohibit or significantly
limit the type of collection and disclosure that would allow for the development of such AI.

Presently, the United States and China are locked in a race as the world’s two competing AI superpowers.
The United States is ahead, but the lead is narrowing. Americans value privacy and enjoy the protections
they are afforded by the Fourth Amendment and other regulations, but the importance of privacy should
be weighed against future economic and national security interests. Where American AI developers are

31 32 33 34 35 36 37 38 39 40 41