A few months back our Think Tank was working through a thought problem, we were trying to consider how best to launch an artificial intelligent software system to run an entire energy grid for a state, nation, or region. Okay so, let’s talk about this.
You see, the entire point of switching over such a system is to take human error out of the loop, and let it run on auto-pilot. When discussing this someone mentioned the challenges of the “change-over” and yes the entire point with AI is to install it and let it do its thing, but you can imagine the hesitation to just let it go from Zero to Mach 2. Thus, they noted; “to go from absolutely nothing to making its own algorithms. Instead of getting some basic start programming done by humans, that’s a scary proposition.”
Indeed, however if we don’t trust the system to work and set itself up, then indeed, we can trust it when there is a catastrophic failure somewhere in the energy grid – when it has to go to work and fix the problem, we can’t just let humans watch what it is doing and intervene, if we do that, we don’t need the system in the first place. Interestingly enough, AI computers which have programmed themselves come up with really interesting program language.
Another problem we noted was that unfortunately these days means too many different things to too many people, and to get the powers that be to go for it, and to allow an Artificial Intelligent software program I to run the command and control for the energy infrastructure will take some rather heavy convincing, and endless congressional committee meetings, not to mention dealing with the power base at the bureaucracy.
A true AI system would go in and figure out what’s there, and test all the systems, and then determine what is needed, I mean I am talking about true AI in the coming years, intelligence far surpassing humankind, personally, and jokingly; that wouldn’t take much in my opinion judging by the latest nonsense in the inquirer and on Facebook. My acquaintance, then stated;
“You mean for example, instead of providing legs and walking as a means of getting from point A to point B, just tell the program to do it. No given basics. I mean legs are probably not the most effective way.”
Indeed, sure, sounds like the Big Dog DARPA concept. That’s kind of how Fred Smith of FEDEX always ran his company, give the instructions; let the special teams figure it out. In the human bio-system “life will find a way” and it will adapt. In AI it will work to complete the tasks given with the tools it has. Philosophically it’s all the same right? Just like the military gives its net-centric components and the field commander’s control, and merely tells them the objectives to perform, or the FAA regulations for “pilot in command” decision making.
If we don’t trust the machines, then we shouldn’t even be discussing giving them control in the first place, which is the over-all fundamental argument with super AI running human civilizations. Please consider all this and think on it.