What can’t AI do? Any time you see the word instantaneously in connection with what an Artificial Intelligence system can do, that should raise a few red flags.
In the case of cybersecurity, this is no different. Countless claims have been made related to the potential of AI revolutionizing cybersecurity as we know it. In truth however, judging by the opinions of Wired and other publications, it can be quite difficult to sift through the hype in these cases.
Because of this, it might help to consider a few key rules of thumb as you read through anything that touts the strengths of AIs. Using Wired’s piece on the connection between AI and Cybersecurity as a guide, we have distilled this discussion down to one key point to set the scene for future analyses.
Judging by the examples in the previously mentioned Wired piece, what companies call AI is often not AI. In other words, it is often simply software that is powered by Machine Learning algorithms. If Wired’s sources are to be believed and they are credible, knowing the difference between this kind of software and full-fledged AI systems is more important than ever.
At this point you might find yourself wondering, how can I even begin to conceptualize this?
In the end, it all comes down to the the fact that ML algorithms are mere cogs in the machines that are AI systems, albeit major ones. Imagine a cybersecurity company that wants to improve the ability of its’ software to detect threats over time. In an overarching sense, one road to achieving this would be to implement ML algorithms that can help the software learn to improve over time. The other would be to basically re-develop the existing software as an AI agent so that it all exists as a self-learning system.
According to Wired, choosing at least one of these options is unavoidable because even back in 2016, firms in the cybersecurity industry were dealing with an average of 200,000 threats per day. Still, you might find yourself wondering at this point: how does any of this directly relate to understanding the true impact that AIs are having on cybersecurity?
In a word, both options take the responsibility of dealing with a number of threats that has become almost insurmountable, away from human teams. With this in mind, one could also reasonably assume that this means their time is freed up for other, perhaps more important endeavors. Circling back to our mention of the difference between AI systems and ML algorithms, in the first case, theoretically no human involvement is needed. While this directly hinges on whether the system is based on supervised or unsupervised learning principles, it especially holds true in comparison to a system that fits the second case.
When only ML algorithms are used, workers with key knowledge of how these functions as well as in how they can and do interact with a network is essential.
As you dive into any sort of reading on these subjects, keep these basic differences in mind and you will be well on your way toward mastering sifting through media hype.
References:
https://www.wired.com/story/ai-machine-learning-cybersecurity/
