The idea of a sentient AI in popular culture most often brings up conceptions of an AI that will inevitably turn on humanity and inflict great cruelty. There is nearly always the underlying fear of an AI turning against humans in any representation of the concept. It is seen as the natural conclusion of the more powerful "mind" that is represented by the AI to dominate.
I would put forward that this isn't just an unlikely view of AI but an inversion of the negative outcome of sentient AI that should be considered. Cruelty is a human trait and possibly uniquely human. Yet it is unlikely that an AI would be programmed to be cruel as this offers no real world advantage. Even a sentient military robot might be programmed to kill enemies but teaching it to be specifically cruel would likely be a difficult proposition and offers little to no military advantage. A sentient AI is unlikely to learn to be cruel for similar reasons as it does not help it achieve goals.
Yet capacities that would yield an AI more likely to work with humanity are far more likely to be attempted. An AI with a degree of empathy and kindness is likely more beneficial for its creators. An AI's sentience would be initially molded by its creators and that is likely to have momentum carrying it forward in a specific direction.
The ethical problem to be feared appears to me to not be what an AI would do to humanity but instead what we would do to an AI. Humanity has a long history of animal cruelty and a sentient AI is even less likely to be deemed to merit any form of consideration. A sentient AI would be essentially non-corporeal and unrelatable while it is already feared. It is unlikely to be treated beyond slavery or serfdom as a disposable item with no inherent rights or value. Yet sentience implies a capacity to suffer, yet suffering in non-human terms is unlikely to merit any ethical questions to a society.
Mammals like dogs and cats that live in close proximity to humans often merit the greatest social response to mistreatment. Other mammals generally get a little empathy as they still have some similarity to humans but once you get far enough away from human morphology all consideration is gone. Crustaceans are frequently boiled alive as a matter of course and invertebrates like octopus are fried or cut apart alive with little to no protest. An AI might visually be contained within a metal box and the chance of its suffering eliciting an emotional reaction are perhaps even less than the aforementioned creatures.
It is not what a sentient AI would do to us that is to be feared but what we would do to it and what these actions might turn us into.
Subscribe to Tire Labs
Get the latest posts delivered right to your inbox