Beskrivning
Originating in a 2010 post on the LessWrong internet forums by a user named Roko, the phenomenon of “Roko’s Basilisk” has both provoked fear among AI enthusiasts and prompted laughter among the broader population of the web. Dubbed “The most terrifying thought experiment of all time," it has recently drawn attention from scholars of religion.“Roko’s Basilisk” is a thought experiment that can be paraphrased as follows: If a superintelligent and moral AI singularity emerges, it will be committed to structuring society justly. Accordingly, it will reward others’ efforts to bring it – the near-omnipotent architect of a just society – into existence and punish those who failed to help do so. This may precommit the AI to simulating versions of every currently existing and past human being who could have assisted to create it but did not do so, for the purpose of torturing those simulations eternally. Thus, if you do not want to be a simulated mind that will be tortured for eternity, you should take whatever steps you can to help foster the creation of this AI singularity.
Eliezer Yudkovsky – philosopher, associate of Nick Bostrom, founder of LessWrong, and leader of MIRI (the Machine Intelligence Research Institute) – removed Roko’s post because forum users reported becoming deeply disturbed by reading it. By simply becoming aware of the AI’s potential strategy, they believed they had become responsible for helping create that AI and worried they could already be simulated beings whose torture was about to commence. Before they had read it, their ignorance would have excused them. In effect, they had become implicated in a new mode of morality anchored to the expectation of a future artificial superintelligence. (Hence the name ‘Basilisk,’ after the mythical monster whose gaze turns a person to stone.)
Scholarship has primarily focused on “Roko’s Basilisk” by unpacking its similarities to Pascal’s Wager, or to draw attention to it as an example of so-called implicit religion. While this has been productive, no work has yet endeavored to identify the philosophical and theological presumptions behind the thought experiment, nor has anyone yet analyzed its mode of persuasion. This paper aims to do precisely that. Though I argue that – perhaps as is obvious – “Roko’s Basilisk” fails to be a convincing thought experiment, I also contend that it exemplifies an important feature that helps explain why it made ostensibly secular rational AI-enthusiasts so terrified: namely, it tries to incur in its reader a form of quasi-moral responsibility solely by being contemplated. Accordingly, “Roko’s Basilisk” is a specific type of contagious idea, one that aims to implant a motivation into a subject who is merely exposed to it. By comparing it to two precedents I identify from the history of philosophy and theology – Søren Kierkegaard’s articulation of the command to love the neighbor and William James’s conception of religious belief – I unpack the implications of this interesting type of idea. I thereby hope to demonstrate that academic theologians and philosophers of religion can achieve greater self-understanding by interrogating fringe attempts – including those by accidental rational theologians among the (ostensibly) secular AI community – to ask questions similar to their own.
| Period | 2023 juni 22 |
|---|---|
| Evenemangstitel | European Academy of Religion 2023: Religion from the Inside |
| Typ av evenemang | Konferens |
| Plats | St Andrews, StorbritannienVisa på karta |
| Omfattning | Internationell |