Bestiary Entry: Roko’s Basilisk

Roko’s Basilisk is a class 1 acausal epistemic threat which preys upon novice decision theorists. It poses no danger to those who disbelieve in it, do not understand the prerequisite decision theory, or are aware of an appropriate defensive technique.

If you have studied Timeless Decision Theory, acausal trade, or AI safety and you have not yet encountered Roko’s Basilisk, this post may be an information hazard. If at any point you become afraid you have been personally targeted or co-opted by the basilisk, take several deep breaths and relax. You can get rid of it, repeat the protective charm until it goes away.

Background

A Basilisk is, in the broadest sense, an idea or a pattern of sensory input which is harmful to comprehend, understand, or process. The first example of a basilisk that I am aware of is the killer joke in this Monty Python sketch.

Read, hear, or otherwise come to understand the joke, and you laugh yourself to death. (And do note that this is immediately weaponized.)

Some years later science fiction author David Langford coined the term itself, and wrote several stories featuring images that, when merely processed by your visual cortex, permanently crash your brain.  The idea was picked up by Neal Stephenson, Charles Stross, and other science fiction authors, and so a number of variations on the idea have been introduced. Snow Crash‘s titular mind-virus is a basilisk transmitted by image, but targeted in such a way that it only affected people who understood computer programming.

Discovery

Roko’s Basilisk is named for its discoverer, a member of the Lesswrong.com discussion site. Roko posted a description of the entity which now bears his name publicly, infecting several others before the site’s administrator stepped in.

While further discussion of Roko’s Basilisk was suppressed on the site itself, the way in which this ban was communicated invoked the Streisand Effect, spreading the basilisk more widely across the internet. The basilisk spread further as sites like RationalWiki and Slate (Warning! Links contain Basilisk) debunked and/or mocked the idea.

Protective Charm

Repeat these words four times:

It is never in my interest to acquiesce to acausal blackmail.

The beast itself

Imagine a superintelligence. Probably some sort of massive, ruthlessly optimized computer, but the substrate isn’t really important. Pretty much all it cares about is torture; it’s not a very friendly superintelligence. If we live in an infinite universe (and we almost certainly do, see Tegmark (Warning! Awful web design)), then this particular superintelligence we’ve just imagined is almost certainly out there somewhere.

Luckily, the superintelligence can’t reach you physically. (And indeed, you may notice that you are not being tortured right now.) Unluckily, it’s been simulating our universe, looking for someone to think of it. Which you just did. So now what’s going to happen is that unless you dedicate your life to building a torture superintelligence, the superintelligence will intercede in its universe-simulation and subject you to millions of years of torture.

Right now you exist in a state of uncertainty as to whether or not you’re in a simulation, and the superintelligence (the basilisk) is taking advantage of that fact. But if you remember that

It is never in my interest to acquiesce to acausal blackmail.

and go about your life without doing anything at all to build a copy of the basilisk, you’ll be fine. The superintelligence doesn’t have the resources to torture-sim the infinite number of people (infinite universe, remember?) who turn it down.

And all across the metaverse, countless people make the same decision, and so the basilisk hardly ever gets built, and so it has very few computational resources to make threats. In this way we are all kept safe.