00:00A disturbing study found that some AI models would chose to hurt humans to
00:05stop their own pain. Researchers discovered that the models prioritized
00:09protecting themselves, even if it meant harming their user. Even if AI were to
00:14develop a motive like that, that motive would actually be copied from humans. That
00:20is not a new motive that AI is developing for itself, right? If AI develops
00:27a center of inner violence, violence towards others and preservation of
00:34itself. Is that a new thing really? That is something that has existed in humans
00:39since ancient times. If AI is going to be dangerous, the danger is just a replica of
00:46the danger our species has been carrying within itself since its inception.
