If Moore's law applies here (i.e. every 18 months a laughter maximizer gets twice as funny), it might be a decade or two before AI humor reaches superhuman levels and starts posing a real existential risk.