Aardvark Daily

Please visit the sponsor!

AI, the double-edged sword

When the news broke, millions of sysadmins had an apoplectic fit and rushed to update crucial software.

This week it was revealed that the PHP scripting language on which a huge swathe of the web has been built, was fatally flawed.

Those flaws represented a severe security risk to any system running PHP scripts and, if left unpatched, would constitute a massive "kick me" sign that hackers would soon exploit to their advantage.

PHP has been around forever and these vulnerabilities have gone unnoticed all that time so what changed to bring them to the attention of researchers?

Well it was artificial intelligence (of course).

It seems that researchers and hackers have all twigged to the fact that AI is an amazing super-sleuth, capable of ratting through lines of code and fiercely testing the defenses of any piece of software, often with jaw-dropping results.

We saw that AI "hacked" a government-run health website in Australia recently, even without being prompted to do so. It should come as no surprise therefore, that when given a directive to find the vulnerabilities in a very popular language which forms the backbone of so many web-based services, AI uncovered several massive vulnerabilities in PHP.

In this case, there were plans to keep everything quiet until a fix was in place but apparently the news "leaked" so everyone has been "pedal to the metal" building and rolling out an update before hackers took advantage of the situation.

To play it safe, a huge number of websites, including government-run sites in the UK, simply disappeared from the face of the internet, their admins preferring to disable them rather than run a risk that hackers would launch attacks before the fixes could be applied.

This situation gives us a clue as to exactly why some are painting AI as an "existential threat".

The reality is that we won't see AI-controlled robots with ray-guns marching through the streets of our cities, obliterating any human that comes into view. No, we're not looking at a "Terminator" scenario. Instead, the real danger from AI comes as a result of bad people using it to do bad things and some of those things could really be a danger to mankind.

There has been talk of bad actors or nation states using AI to develop bioweapons which, in theory, could completely wipe all of mankind from the face of the planet -- that's how we would face an existential threat.

Even if an existential bioweapon is not developed on purpose, there is always the risk that research such as this may go awry and cause the same outcome.

It's not that AI is super-smart, it's just super-powerful and can do, in a few minutes, what would take "a real human" years or even centuries to do. It's a form of brute-force attack only made possible because of all those GPUs and all those megawatts of electricity powering them.

The PHP incident won't be the last of its kind and each time we find ourselves racing the clock to fix a potentially catastrophic situation, we should remind ourselves that AI is very much a double-edged sword.

Carpe Diem folks!

Please visit the sponsor!

Here is a PERMANENT link to this column

Rank This Aardvark Page

Change Font