The Blog

CyberWinter Is Coming!

We are today faced with a grave danger posed by the rise of the machines. Between today and 2022, according to different but converging estimates, one supercomputer somewhere in a civilian lab will reach the theoretical calculation power of the human brain.

Games of Thrones, the TV series on HBO, starts in the first minute of its first episode with a very menacing threat, that of the 'White Walkers', ghosts from the deep north and from beyond the winter of life. But then, we lose sight of this mythical menace. However, it remains in the background whenever someone is saying "Winter is Coming!". We see different type of dangers, and we forget this initial warning, lost in other myths and petty conflicts... And then it hits us five years later, at season five, episode eight - and we understand that, indeed, that very peculiar "Winter" is the worst thing that could happen to the earth.

We are today faced with a grave danger posed by the rise of the machines. Between today and 2022, according to different but converging estimates, one supercomputer somewhere in a civilian lab will reach the theoretical calculation power of the human brain. Some actually put it at around 10^16 floating operations per second. If that's the case, we've already crossed that threshold last year. More importantly, it takes usually less than 20 years for the calculation power of the top lab to get in the computers found in mainstream commerce.

Imagine the type of power we'll have at our finger tips in 20 years from now. To add to that - some US Air Force researchers estimate that we may reach two trillion connected devices by 2035. If only 5% of them are autonomously moving - then there'll be more robots than human beings on earth in 20 years from now. Based on the observation of current robot population and its dynamics, others arrive at the same conclusion.

But beware of the wrong myths, too. Very smart people like Elon Musk or Stephen Hawking have been warning us about the threat of IA. Some of their warnings are misguided or imprecise. Machines are not going to seize power once they become super-smart to get rid of the human species. Cognition and emotion are two separate brain functions. Our emotional system is shared in large part with all our mammal brethren, as neuro-scientist Jan Panksepp has demonstrated. The emotional system, and the reactions it triggers, was developed for one specific reason: reproduction and survival. But machines are different. They are not going to fight for the girl next girl - they are not vying for reproduction. They are not even fighting for survival. By design, they can be switched on and off and on again. They can go from life to death and back to life at the user's will.

The actual problem does not lie with the machine but more precisely with software, the digital language that human user and machine use to communicate one to the other. This is also what is called "cyberspace", hence the "cyber" threat. There are still no methodologies to this day to release software without zero bugs. Even if there was one, because that new software would interconnect with all the other software layers in Cyberspace, it would still be potentially compromised by the most vulnerabilty-riddled software it interacts with.

When these vulnerabilities are actioned by malevolent human users, they pose all sorts of catastrophic dangers. With Internet of Things and mass robotization on its way, we are going to live soon in a fully programmable world. But in a programmable world, it is easy to re-program a tool into a weapon. So, all of a sudden you lose the control of your car - this was demonstrated in 2015 by researchers able to take over remotely a Jeep Cherokee. This explains why cybercrime may rapidly expand and reach 2-3 times the amount of money made by the illegal drug traffic markets. Besides, thousands of cars could be taken over the same way, and crash the same day. This could give ideas to terrorists.

Behind lie all sorts of grave dangers. We could have the rise of "cybercrime-states" as we had "narco-states" - states corrupted to the core by the money in cybercrime. They could become safe haven for black hat hackers, mercenaries working for other rogue states or cybercrime empires. But if we cannot control our programmable world, then global economy will collapse - a threat mentioned by Director of National Intelligence James Clapper in hearings to Congress in 2015. Then - terrorists could create mass terrorism waves. Even worse: programmable military systems could be hacked. This has already happened in the past. This risk is the reason why US nuclear missile silos are still using 1980s computer equipment - to avoid the vulnerabilities that may come with a major IT upgrade. If whole classes of conventional weapons are becoming "hackable", armies are going to become very nervous.

Add to that the very risk already posed in terms of international stability by cyber-weapons. They have no doctrine of use today as clear as that of nuclear forces during the cold war. However, the sabotage of the Natanz uranium enrichment plant in Iran in 2007-2010 by a cyberweapon has demonstrated their strategic potential. This adds to an escalation risk, already taking place in slow motion.

And this all create a very dangerous mix that could reach its acme as we achieve a fully programmable world by the middle of this century - only 30-35 years from now. That is way earlier than the risks pose to earth by global warming. And the funny thing: very few people talk about it.

Indeed, beware. CyberWinter is coming.

Before You Go