July 18, 2017 —
I am using that well-known adage to start this missive as a lead in to talk about getting one’s arms around the development of artificial intelligence. Most of the noise around AI is that it will be the great enabler – from 5G to the Internet of Everything (IoX). But, like all evolving technologies there are bumps.
Hype aside, AI is a two-headed monster. If there is anything history has taught us it is that as powerful as a tool can be, that power can be subverted to do evil.
One of the most visible vectors where AI holds a lot of promise is cybersecurity. But this vector also has some of the highest potential for fallout. This because the same approach that is applied for cybersecurity is used by the dark side to circumvent that security. In fact, sometimes it is easier for the hacker to break security using AI than it is AI to set up a good security screen.
Why? Well, because, truthfully, the code is imperfect. It is usually much more complex than it needs to be, riddled with a variety of flaws and usually a concoction of multiple programming methodologies and programmers. These elements make code a wide and shallow target for compromise. This concern becomes vital with 5G and Internet of the IoX devices, many of which will operate autonomously. Others will be in the hands of the consumer, who has little comprehension about the metrics of cybersecurity.
Another AI dark side capability is that it can be made to mimic human behaviors. An unsuspecting user might be tricked into revealing a password or allowing the mimic AI to allow access to a mobile device, for example. It might also direct a user to download malware files, make a shady transaction, or relay confidential information. It could be capable of doing the same for autonomous devices.
These are only two of many potential cracks in the AI ecosystem. As AI evolves, there certainly will be more, some of which will surprise us.
Now, moving to the light side, one of AI’s strongest assets is in cybersecurity to analyze code. With modern high-power hardware, AI can run though lines of code much faster and more effectively than both humans and present analysis software. It can root out flaws in the alpha and beta stages before the application or program goes public. That closes a significant window of opportunity for hackers since the code now leaves the “factory” with a much higher percentage of un-exploitable programming. AI will also be capable of monitoring such devices and analyze existing code that is already out there; and patch that code as well. This closes a ubiquitous window of opportunity for hackers.
Another rising star in the AI wheelhouse is unstructured data. Present day hardware is very good at processing structured data. Unfortunately, the universe is not made of that. Real-life data is unstructured, which cannot be effectively analyzed without a lot of massaging first.
This vector has a lot of potential. AI can be used to analyze unstructured data and learn from the non-linearities that it contains. It is the content that is the most valuable. For example, AI can be used to understand how the human brain makes non-logical assessments. Something that is difficult for today’s computers to comprehend.
This can be extremely valuable in the mobile world – autonomous vehicles, smart “X,” consumer retail, and enterprise apps, just to mention a few. All of these ecosystems have unique anomalies that make data in them unstructured. Couple that with Big Data and having a fuzzy approach to analysis provides much more reliable and specific results.
AI is still just emerging. There are some novel examples out there; the most visible is in human-centric robots. While that is an attention-getting vector, the real AI will be much more of a behind-the-scenes platform.