I’m not sure if there is any difference. They are assuming that their “AI” is acting alone and in a very narrowly defined context.
What’s to say the human voluntarily entered this capsule and the “AI” does what that person wishes? One could take the Kim Jong-nam assassination as an example where the participants may have believed something completely different was occurring than the true context.
In the absence of “AI” someone could reverse the do die/do not die button. However, adding an additional, pliable layer is a bad idea.
Whether it is autonomous weapons or this, once AI or machine learning is added, the human is no longer in full control.
What’s to say the human voluntarily entered this capsule and the “AI” does what that person wishes? One could take the Kim Jong-nam assassination as an example where the participants may have believed something completely different was occurring than the true context.
In the absence of “AI” someone could reverse the do die/do not die button. However, adding an additional, pliable layer is a bad idea.
Whether it is autonomous weapons or this, once AI or machine learning is added, the human is no longer in full control.