The entire point of the 3 laws of robotics is to show how it's pretty much useless to come up with rules of robotics.
They are not meant as a practical guide.
The only way to a robot to follow the laws of robotics is for an entity to be able to identify the intent of the three laws. This implies human level reasoning at the least. Further, in order to obey the intent of the laws a robot has to be able to deviate from the literal interpretation.
So... in the end any system of hard logic will fail. Having the capability of fuzzy logic decisions and the ability to re-work the framework is necessary.
Light's solution is the better, even if it allows for exploits. All systems are exploitable.
They are not meant as a practical guide.
The only way to a robot to follow the laws of robotics is for an entity to be able to identify the intent of the three laws. This implies human level reasoning at the least. Further, in order to obey the intent of the laws a robot has to be able to deviate from the literal interpretation.
So... in the end any system of hard logic will fail. Having the capability of fuzzy logic decisions and the ability to re-work the framework is necessary.
Light's solution is the better, even if it allows for exploits. All systems are exploitable.