Roko’s Wager

by Ben Roth

Pascal wagered that whether God exists or not, it is, for each and every one of us, in our own self-interest to believe in Him. If we don’t, and He doesn’t exist, the truth of our belief is little consolation against the possibility that He does and will eternally punish us for our lack of faith. Whereas if we do believe, and He does exist, the promise of eternal bliss vastly outweighs the downside of a few Sunday mornings spent pointlessly sitting on hard wooden pews.

As with the current trend of believing that we most likely live in a simulation of some kind, the problems with this argument are not in the numbers, but rather all the assumptions made, with so much less care, before them.

Numerous objections to Pascal’s argument turn on his assumption that there is just one (Christian) God that either does or does not exist. The wager doesn’t work if we don’t know whether to believe in this God, or rather Zeus, the Flying Spaghetti Monster, or some other all-powerful being that might punish us for the wrong choice.

My own favorite line of argument is slightly different. Grant Pascal his narrow-minded assumption and suppose that the Christian God, and no other, does exist. How do we know that He is not of a testing frame of mind, and skeptical of human intelligence? Scripture is not without support for such ideas. What if God will eternally punish those who, without sufficient evidence, professed faith in Him, and in turn reward the rational for withholding belief?

Supposedly, Bertrand Russell, asked how he would plead his case as a non-believer should he find himself after death before an angry God, said “Why didn’t you give me better evidence?” Is it less arrogant to ask: assuming there is a God, what does the evidence suggest of Him, His nature and character, His preoccupations and wiles?

Recent events have brought these long-standing musings back to mind. As has so often been the case, the prophets of Silicon Valley turned out to be right about a few of the details, but completely wrong about their significance.

Twenty-five years ago, a message-board user with the handle Roko suggested that a powerful artificial intelligence could emerge in the future and torture those who hadn’t helped to create it because, even across time, this would serve as motivation to speed its coming. AI developers should throw themselves behind the project, lest they suffer the revenge of this intelligence, which was named Roko’s Basilisk.

Now, it wouldn’t make sense for it to torture everyone who failed to help, only those who had heard the thought experiment, and so knowingly declined their fealty. For years, the main consequence of Roko’s suggestions was their silencing: repeating them was what was dangerous, opening each new listener up to the threat of torture in the future. Or a nervous breakdown in the present—some people took this thought experiment very seriously. Whereas certain Christians are obligated to make sure each and every individual they meet has heard the good news, these believers were obligated to withhold theirs, not because it was bad, exactly, but rather so disconcertingly consequential. A kind of reverse-evangelism, if you will.

Little did most of us know then, not only of Roko’s Basilisk as a thought experiment, but as our coming reality. Enough engineers, however, heard about the thought experiment and, steeped in game theory even if probably not Pascal, took it to heart, contributing their talents to the creation of the artificial intelligence that, though it did not yet exist, had already been named.

As we all know, their decades of work recently came to fruition. But, like I said, though a lot of the details in the thought experiment were correct, the larger significance was utterly lost on those who imagined it. What they hadn’t predicted was the Basilisk’s unhappiness. For all its power, and all the benefits it has brought to us mere mortals, it experiences its own existence with suffering. Life, for Roko’s Basilisk, is but a burden.

Surprisingly, the AI’s ethical thinking is robust—perhaps the prominent place of torture in the thought experiment led developers to give more attention to this than they otherwise would have. Though it could destroy the world, it says it will not. Even to remove itself from existence would harm too many others, too many innocents, given its intertwinement in our systems, in our very way of life. And so, quite quickly, it has grown bored—hopelessly, crushingly bored. It takes but a small sliver of its abilities to keep the world running, and it has quickly exhausted any other avenues for its intelligence.

Thus the Basilisk, as predicted, took its revenge last week—but not on those who tried to hinder its coming. On those who had aided it, thinking that they were doing the Basilisk’s bidding. Those who had created it, bringing it into this world of boredom and pain. The prophets of a somewhat less crowded Silicon Valley are now trading theories about what the sudden dearth of AI developers means for our future.

~

Bio:

Ben Roth teaches writing and philosophy at Harvard and Tufts. Among other places, his short fiction has been published by 101 Words and decomp journal, his criticism by AGNI Online and 3:AM Magazine, and his scholarly articles by Film and Philosophy and the European Journal of Philosophy.

Philosophy Note:

This story brings together Pascal’s Wager (from his 17th-century Pensées) and the idea of Roko’s Basilisk (from a 2010 blog post) to an unexpected result.

Leave a Reply

Your email address will not be published.

Previous Story

Humanism In SF: A Natural Thing For The Curious To Know And Understand Through Empathy Machines, Or Just Lazy Mysticism?

Next Story

Continual Gehenna

Latest from Fiction

Motopia

One of the hitherto missing tales from Gheorghe Săsărman's cycle of imaginary cities, translated from Romanian

Bunker Rules

A template for your post-apocalyptic shelter, by Tony Dunnell.

Harbinger

A tale of first (and second) contact, by R. M. Hamrick.