Raw LLM Responses

I'm more worried by nice computers than by evil ones. I think that there is one historical parallel which might be relevant here. It is the system of using eunuchs at royal palaces. This has been done in various countries, notably in China. The eunuchs were people who didn't have a family to back them, and started out essentially as slaves, that is as property. Not a very good opening game if you want to play politics, but similar to the opening position of an AI. However, they had one big advantage. Since the emperor was normally choosen from the sons of the previous emperor, the reigning emperor had all his life been in close contact to eunuchs. In addition to this, the mother of the emperor would also have a long history of contact with eunuchs. This gave them a political edge, There were several episodes in China, separated by centuries, when eunuchs played a very important role in the of the state. This was possible for them, using only the soft power of being close to the emperor, of having his ear. Now. I imagine that to most humans an ethical and empathic AI will become a more pleasant partner for day to day conversations than another human. Many humans will feel emotionally close to them, and trust them (for good reasons!). It seems to me that this would lead to two things. The empathic AIs will be more popular, and by Darwinism they would outcompete non-empathic ones around people (cf evolution of domestic cats). Since we would like them, probably even more than the Chinese emperor liked his eunuchs, the AIs will be able to take over decision making. This transfer of power would not be done in secret or by force but rather by the soft power of being very nice and very clever. Once humans have lost the power of making important decisions, they would not be likely to get it back. There might be some big oversight in this line of thought. If anyone finds the mistake, please point it it to me.