I've seen numerous different discussions online about how to handle artificial intelligence with regards to moral laws. How should we, as orthodox Jews, dispatch of these problems? I turn to this law for guidance,
Exodus 21:28 "When an ox gores a man or a woman to death, the ox shall be stoned, and its flesh shall not be eaten, but the owner of the ox shall not be liable."
The way I read that needn't be specific to an ox. An ox is just one being with agency. Here the law calls for immunity for the owner of the ox, and the stoning of the ox (the subject of the crime). No punishment is dealt to the parental oxens or the species. In the case where a machine injures another, should we turn to Exodus 21:28? If we did, we would grant immunity to the owners and creators of the machine and stone the machine? This may be more difficult to contend with if we have to punish The Cloud, but given enough time and stones I don't believe it's impossible.
The full context of this can be found here,
28 “When an ox gores a man or a woman to death, the ox shall be stoned, and its flesh shall not be eaten, but the owner of the ox shall not be liable. 29 But if the ox has been accustomed to gore in the past, and its owner has been warned but has not kept it in, and it kills a man or a woman, the ox shall be stoned, and its owner also shall be put to death. 30 If a ransom is imposed on him, then he shall give for the redemption of his life whatever is imposed on him. 31 If it gores a man’s son or daughter, he shall be dealt with according to this same rule. 32 If the ox gores a slave, male or female, the owner shall give to their master thirty shekelse of silver, and the ox shall be stoned.
Another fellow Jew commented on the relevance of Exodus 21:29. I believe it is also applicable however to a much lesser degree. Exodus 21:29 requires
- A pattern of goring in the past. For our analogy, this would mean a habitually malicious AI.
- Its owner has been warned. This seems totally reasonable. I imagine a lot of AI owners will be warned, and have this criteria satisfied.
- The failure of the ox to be kept in. I read this to mean the owner having reacted to the warning to mitigate future harm. It seems unlikely that an owner of a malicious AI will be so negligent as to ignore warnings.