You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
Weaponized AI is here and more is coming, we just haven't had much exposure to it yet. It can take forms we haven't even imagined and dangerous persuasion is just the first subtle ripple in a coming tidal wave. Weaponized AI will be programed to hibernate, hide, maneuver, attack and retreat all in stealth. It is already lying in wait somewhere right now, just waiting for the right trigger. Even if discovered we will not hear about it for to tip ones hand to a method of discovery will give its creators more power to make it even more dangerous. Forms of weaponized AI will extend from the simple made by an individual to attack another individual to that built by nation states to attack other nation states. AI war will be the ultimate chess game and will rapidly join the ranks of sophisticated nuclear/biological/chemical warfare. AI will directed at the weakest targets first and the options here are endless. The foundation of any defense will be based on careful design of semis at all levels. The days of building semis without taking this into account will have to come to an end shortly and should be one of the primary considerations of the design of any EDA program. Just like a standard physical military defense of many layers, the same system will have to apply to the hardware of the virtual world. A wide variety of defenses will greatly reduce the chances of disaster for an aggressor will have less of a chance if faced with many unknown, hidden and strategically creative defenses. An ounce of strategy will be worth a ton of hardware. Secure logic chains that can't be broken without destroying them I feel will be a key part of the answer to this current and looming problem. Circuit breakers of various forms will be our best defense and will sharply limit the damage while raising alarms. Any thoughts, comments or additions welcome and solicited.
I see it coming as well but I seriously doubt there is anything we can do about it. Think of all the cyber crime that we now see on a regular basis. There is just no stopping it. AI will be weaponized in the same way by governments (military) and criminals. It has been interesting to see science fiction come to life but I am not looking forward to seeing the Terminator movie or Skynet come to life.
Dan, As the Equifax demonstrates, there should be a system where if updates aren't done, the system shuts down. Also, security will have to be a combination of hardware and software at all levels in both. Security, transparency and a logic chain need to be maintained in everything. Also, I don't sign up for everything and don't use a debit card for a soda. I tell my wife, every card or open account is another door to open and the more doors, the weaker the system. I like Oracle's new cloud management system where all updates are automatically taken care of.
Managing the particular AI for your field of endeavor is the future. The semi and full automation of everything is coming faster than most realize. This has been the foundation of how I make a living. Certain areas that are unexpected and not obvious are the big winners that have to rotated depending on their projected life span. Some companies have mastered continuously changing evolution methods, among them AMAT and TSMC and they are basically partners as Morris Chang has talked about this strategy many years ago.
