Artificial intelligence (AI) is underpinning many striking new developments. We have self-driving cars with predictive abilities, voice-powered personal assistants, behavioral algorithms that can predict what we want to buy online, smart ‘agents’ that anticipate what movies we want to watch and so on.

It’s hardly surprising then that business leaders want to use AI to drive transformational advancement.

Companies and executives that fail to do so risk being left behind.

But what many companies are calling AI today is an algorithm that responds based on pre-defined inputs or user behavior. This isn’t really AI. A true artificially-intelligent system is one that can learn on its own.

What we have today is machine learning, which is a subset of AI. That said, it is certainly driving tangible benefits to organizations and many executives can articulate the high-level benefits of this type of AI.

The bigger issue, however, is whether you understand the deeper promise and limitations of AI in your particular field and are you able to separate the hype from the reality?

To do so may well require a more practical working knowledge of how to use machine learning technologies as well as acknowledging its potentials and limitations.

Promise, potential and limitations of machine learning in cyber security

Before we get to this let’s look at the promise of machine learning and its potential limitations in the field of information security industry which are equally applicable across many industries.

Hundreds of companies are already incorporating artificial intelligence into their technologies to predict, prevent, and defeat the next major cyber-attack.

For instance, automation and machine learning can analyze the normal behavior of privileged users, privileged accounts, privileged access to machines and authentication attempts.

It can then identify deviations from normal profiles and flag them as potential cyber security breaches.

Machine learning algorithms that continually adjust the baseline mean you can continually adapt to a changing risk environment. For instance, when a new attack vector is discovered the algorithms can be adjusted to factor this in to analysis.

Insights into evolving cyber threats

At one level machine learning can stop us from becoming too overwhelmed by the rise in the number of attacks; something like 200,000 new malware variants are discovered each day, and this is a conservative estimate.

Machine learning helps solve this volume issue. It can also be used to combine insights gathered from customer data and produce a more complete and immediate understanding of evolving threats.

Recently computer scientists from the UK’s Cambridge University used machine learning to crunch masses of data from 113 known cyber-attackers and built algorithms to compare the data to thousands of underground forum users.

Some 30 million posts from hack forums were processed, looking for key words such as DDoS (a denial of service attack) or people who discussed distributing malware and account cracking.

The scientists were able to pinpoint technical jargon and online idiosyncrasies which suggested criminal intentions.

They combined this with data on the users’ popularity on the forum, ultimately identifying a network of closely connected ‘actors’ exhibiting signs that they could be plotting serious attacks.

The scientists said: “These tools helped to identify user accounts that might require further investigation by law enforcement and security firms monitoring underground communities.”

Glaring ethical issues

Of course, there is a glaring ethical issue. Can you arrest and charge someone for discussing cybercrime on a forum? But if we put this considerable moral issue to one side for the moment, we can clearly see the potential of machine learning to thwart cybercrime.

Unfortunately, and this is an important point, cyber criminals are also finding intelligent new ways to use machine learning to their advantage.

For example, some ransomware attacks have been discovered to use machine learning to get smarter about what information has been encrypted on a victim’s computer and how much to charge for it.

This dichotomy is not just particular to the information security industry it can also be an issue in other industries. We can summarize it as the conflict between potential and limitations.

Deep understanding

Those at the head of organizations would do well to understand these potentials and limitations.

AI and machine learning aren’t a magic panacea but they do offer tremendous benefits.

So, what’s the best way for a C-suite executive to get a deep understanding of machine learning?

For years we’ve taught up-and-coming leaders that programming is the new literacy. While programming has become the lingua franca of digital disruption, we are rapidly moving to a period in time in which machine learning is our new literacy.

It will define and shape businesses and industries and perhaps in even more profound ways than the internet.

If you want to lead an organization into this new era it’s critical you become literate in the language that will matter most.

For this reason, a functional working knowledge of machine learning should be considered a must-have skill for those who truly want to shine in the C-suite.

Fun and stimulating

Don’t be dismayed. The study of machine learning is surprisingly fun and intellectually stimulating.

It’s also invaluable in achieving a deeper understanding of the possibilities and limitations both in general and for your business. For this reason, leaders should place significant value on cultivating a hands-on knowledge of machine learning tools.

The advent of cheap computing resources, vast amounts of publicly available data and the availability of extremely powerful, free and open-sourcemachine learning tools, and powerful libraries that abstract away the complex underlying math, make it feasible for almost anyone with just a little technical orientation, discipline and time to get to grips with machine learning.

As the CEO of a technology company, I’ve delved into the world of machine learning starting with simpler machine learning implementations and then dipping into the fascinating world of deep learning and neural networks.

Informed decision making

While I’m certainly not writing code that will find its way into production, this “hobby” work has had three important benefits: I can converse at a much deeper level with our data scientists and cyber researchers who are building our machine learning models; it gives me the ability to better assess the reality behind competitor and partner claims and I’ve used my machine learning skills to participate in our company’s bi-annual hackathon.

Taken together these points have enabled me:

  • To make more informed management and investment decisions
  • Be better able to separate hype from reality
  • Send a signal to the company that if even the CEO can code a neural network to detect malware anomalies, then we should all be lifting our sights toward the art of the possible

Leadership and innovation

Data science is becoming a must-have component in business. The ability to take existing data that is not necessarily useful on its own and combine it with other data points to generate insights an organization can use to learn more about its customers is increasingly important.

As such the impact of data science is growing and business leaders are going to need a deep understanding of its potential to ensure their organizations can profitably leverage its value.

That said, machine learning is equally important. In some senses they come together given that data science at a core level brings together computational and statistical skills and machine learning for data-driven problem solving.

With this in mind it would certainly be advantageous for executives to develop machine learning skills, even at a basic level.

It adds to important leadership skills that will help steer the organization in a direction that places it at the forefront of new technologies and several steps ahead of competitors.

Blog article originally published on AlgotithmX Lab.