Cyberjester said:
Benjamin Moore said:
Mines a two-parter...
AI head-of-state, with a parliament consisting of non-professional politicians.
...
Everyone regardless of their knowledge has one vote; if they read or watch something about the issue they get a heavier vote; investigating both sides of the issue gives an even heavier vote; all the way to doing a PhD on the topic, or higher.
No point complicating matters, your two-parter is a communist society with capitalist markets and an AI head of state.
Brilliant, take all the good bits you think exist and throw them into an erroneousness system built on a lie.
That being the AI. AI is only as good as the person who wrote it. People are flawed, ergo the AI will be flawed. It's an inescapable fact. Also, I have a virus, head of state nukes China, your world becomes an rl Defcon. Great game. ^ ^
-- snip --
Your response is pretty much the response I expected from my suggestion. Not that you are wrong; your analysis on why there is no ideal government lines up almost exactly along why I suggested the above scenario in the first place.
But your dismissal of my suggestion hinges on two points: AI will never be 'perfect', and no system is completely impervious to attack. The latter is actually false: some military operating systems are mathematically 'proved' to be unhackable, although designing an operating system this way is extremely expensive and difficult to do. But it can be done. Look up 'sel4' from Open Kernel Labs for a downloadable, non-commercial,
formally verified operating system.
The AI claim, that because humans are fallible, so inevitably, will their creations, to me is a defeatist fallacy. It's like the claim that no machine can build something more accurate than the parts it is made up of. This is blatantly false; if it were true, no machine could ever build anything more accurately than we could. We are, after all, machines in our own right. (The problem with that statement is that it is backwards: it should say: nothing can be built that can
machine something more accurately than itself. There are plenty of construction methods other than machining that can, particularly casting and forging.)
But we can design things that are less fallible than we are. Since we are on the topic of AI, lets look at the control engineering side of things, which is what I am. Control Theory is the method of mathematically controlling a system's (plant) response with respect to its input, and usually it's output too (feedback). Such a systems response is highly more accurate than sitting a human behind a control panel and letting them control the device.
But apart from that, as long as the operating system for the AI is sound, the fallibility of the AI is irrelevant. The whole point of the system is the emphasis on individual democracy and transparency. No need for 'deals' because policy goes through everyone. Besides, the fallibility of a government is only with respect to those it governs, and those it communicates with on an equal level. As Cyberjester pointed out, humans are fallible, and so are all other governments. Hence neither national nor international decisions have to be perfect; they just have to cause the minimum amount of damage... I even left in a back-door: the original non-professional parliament is designed to be able to take up power in the event of a breakdown to the system, of which can be shutdown through the ombudsman's office, who cannot take any power.
Besides, even if such a system were possible, even if I was 100% convinced that my system was infallible, I still would net let such a system have 100% command over strategic military assets like intercontinental ballistic (or targeted) missiles. I doubt neither would the average citizen of the state. And that is important, because whether or not it
had control over such assets is a political issue.
tl;dr: some OSs, such as
sel4 are formally verified and are unhackable. The emphasis for the AI is on transparency, not infallibility. Whether or not it is fallible is irrelevant, as no other government system is infallible. It just has to be good enough. Any problems to the system can result in an emergency shutdown, after which elected officials take emergency power until a solution is found.