Law enforcement by AI
From Wise Nano
The model was developed under the assumption that the current law enforcement system is not suited for a society in which an offence can have massive consequences like the death of thousands, with little effort required and likely impunity through anonymity for the perpetrator.
The basic proposal is to implant a watchdog into each person. The watchdog would have to include a very sophisticated AI, doing three things: 1) monitor their environment through the person's senses; 2) compute whether they were about to commit a crime; 3) partially paralyze them in a way that prevented the crime from being committed.
Aside from the likely opposition to having an implanted computer capable of paralyzing you, it's not clear that any AI could be designed that would be able to comprehend whether a crime was in progress. Nanotechnology would not be much help in building such a thing; the proposal is included because of the possible extreme risk posed by advanced nanotechnology (possibly requiring a ver powerful form of law enforcement) rather than because nanotech enables it.
The discussion on this topic is ongoing. Below is the current state of the article/proposal. Some people support it, and others think it is unworkable. Click Talk:Law_enforcement_by_AI to see and join the discussion.
Current interim state of the proposal
In order to prevent crimes from happening, every citizen has a removable device implanted into his spine, replacing the top or second to top cervical.
This device allows access to a knowledge database (comparable to an online encyclopedia) and instant communication with all other members of the network. Also, a yet unknown type of AI system has access to the sensory input of each person and has the possibility to temporarily disable the person´s movement ability to prevent a crime from happening.
As far as we know today, sensory input of all known, here relevant senses (sight, touch and hearing being the most important) is relayed to the brain for processing via neurons, the signals being electric action potentials along the axons. These are easy to measure, especially with microscopic or nanoscopic devices made by a nanofactory. "Easy to measure" also includes scale, because producing large quantities of identical, low-tolerance devices is no problem with a nanofactory. Also, with the device replacing a cervical, the device´s receptors and effectors can be concentrated there as opposed to being dispersed over the whole body and connected through many meters of wire. So the next step is to determine what nerves to tap to get the right input/output. Once medical nanodevices are available for practical use (which is the assumption), it should be possible to reliably identify what nerve transfers signals for what purpose, and configuring the device accordingly.
The AI system interpretes the input and and decides whether a certain action might result in a crime happening. If that happens, the AI can temporarily disable certain motoneurons that control the movement of the peron´s extremities, for example in order to prevent the person from pulling the trigger of a lethal weapon aimed at another person. Simple AI systems might apply the acquired sensory input to a biomechanical model of the human body and heuristic algorithms to predict the most probable action. For example, if a pistol is aimed at another person, then bending the trigger finger will result in the weapon being fired. Once the AI recognizes such an neural order is sent via the nerve system, the AI will prevent the execution by disabling the targeted motoneurons.
It is a question whether the computer system´s hardware will be able to perform the necessary computational steps fast enough to be effective. The computation time depends on the computational speed and available memory of the computer system, the detail level of the applied model, and the desired certainty of conclusion. Calculations in Nanosystems show computation efficiencies of 10^16 instructions/wattsecond. Estimates hold the raw computational power of the brain between 10^13 and 10^16 operations per second, so a 10 watt device should be able to calculate faster than the human brain. Also, at >99% efficiency, the heat dissipated by the device could be handled by the human body, as opposed to requiring additional, artificial heat ablation.
The other part of computation is software. As Virtual Reality and the commonly used Motion Capture method have shown, it is possible to translate body movement into digital data. A relatively simple biomechanical model should suffice to simulate the current movement with the software. Information about muscle tension (which is closely related to the signal frequency of the controlling motoneurons) should allow the software to make sufficiently correct predictions about the immediate future with respect to posture of the person and its relation to the environment.
One possibility to power the computer would be using energy-rich, biogenous molecules like ATP. Assume a peak power consumption of 10 Watts, which results in a 240 Wh energy consumption for 24 hours a day under full load. Assume further that the device goes into stand-by during sleep hours, which we assume at 7 hours/day on average, with negligible power consumption during sleep. Thus we can calculate the average maximum energy consumption at 170 Wh/day. A human body requires about 2500 kcal/day, which is about 120 Watt [1]. Depending on the type of activity (sitting, walking, carrying pianos) a typical human being consumes an estimated 100 to 150 Watt on a typical day to function properly over a long time (months). This means on average he requires 2400 to 3600 Wh/day. Thus the average maximum energy consumption of the device would increase the natural human energy consumption per day by 5% to 7%, which seems to be possible by slightly higher food intake than required for the type of individual activity. This implies the possibility of over-sizing the device to allow for computationally expensive tasks, scaling up the prediction model to allow predicting less obvious action, or using the higher energy consumption to prevent diseases related to obesity, by preventing over-weight. If the computational requirements should prove higher than can practically implanted in a human body, then a solution might be to outsource computation to a mainframe, whose computational power can, for this purpose, be considered unlimited. The connection can be made safe by quantum cryptographical or other methods.
One might note that this approach is open for the development of true AI - that is, a system that is so complex and hard to predict and that can evolve its problem-solving strategies that it could be considered truly intelligent. Of course, then the potential problems that Runaway AI can bring will have to be considered. However, for earlier systems such high levels of sophistication are not mandatory, as outlined above. Also, even super-human AI can´t do any harm if it physically can´t reconfigure its hardware. The worst a malprogrammed, hacked, or runaway AI can do is to disable motoneurons of arms and legs, which is in all but the rarest cases not lethal. Such out-of-control devices can be disabled externally and removed surgically, if they are designed correctly.
Definitions of crimes to be prevented by the AI should be open to modification after democratic process, however, to prevent abuse of the system, basic directives must remain permanent. Permanent directives might include the restriction to capital crimes, i.e. crimes which have the direct consequence of other people being physically injured or killed. This extends to constructing or releasing autonomous devices that can be confirmed to be able to injure or kill other people. The democratic process required to change the definitions need not differ much from the system in today´s democracies. Permanent directives may only be changed if a certain fraction (>66%) of the legislative body and of the population agrees, while minor changes need only be ratified in the legislative body.
One crucial characteristic of this system is its voluntariness. Every member of the society must at any point be allowed to leave the system. To prevent profileration of potentially dangerous designs or enabling technologies, no disapproved designs or products/prototypes thereof may be exported.
All citizens are allowed to submit designs for evaluation by a technical panel, supported by the AI system. If the evaluation sees no reason to dismiss a design, then it can be freely compiled and used by all citizens.
The most important reason to dismiss a design would be its ability to allow a crime that cannot be prevented by the AI, be it through technical or legal impossibility.
