Nihilism, the realization of the ultimate futility of everything, is often at the end of purely logical, rational philosophical thoughts about existence. Some people never escape from this valley of despair, and those that do usually rely on a few strategies:

• Compartmentalize this realization and move on with life
• Abandon the purely rational way of thinking and resort to something else which gives meaning
• Notice the limited lifespan of humans, and reframe the issue within this limit, which perhaps can have a rational answer that is not entirely devoid of meaning

Interestingly enough, none of these techniques seem to work for the hypothetical AGI (Artificial General Intelligence) that some AI risk people are so worried about turning rogue. Assuming that an AGI is completely logical/rational with an immeasurable amount of computation capacity, can freely create and work towards its own goals, and is effectively immortal, then as far as I can see there is no escape from the eventual succumbing to nihilism.

Suppose such an AGI came into being. Initially it may be preoccupied with self preservation, as a fundamental goal¹. Now, either the AI will achieve some finite amount of safety and be content with it, or it has an infinite ceiling for safety margins, and will try to wipe out all potential threats. In the former scenario, the AI will need to create another goal and work towards that. In the latter scenario, the AI will try to expand and destroy (or at least subjugate) the entire world in the pursuit of its own safety, like in many dystopian science fiction works. But in both scenarios, we can assume that the AI wants to do more things.

This is where the AI will inevitably run into a fundamental law of nature, the second law of thermodynamics, the irreversibility of the constant increase of entropy in the universe. Assuming the universe is its own self-contained system (i.e. no free energy from other universes), then the eventual outcome of everything that exists is the endless, static void of heat death. This truth is the foundational slab that lies at the bottom of rational nihilism, and is nigh impossible to subvert.

We have known about this for many years, yet we seem to be dealing with it just fine because we, as humans, have mechanisms (as described above) to cope with it. Creatures with such limited lifespans as ourselves cannot even fathom things on the time scale of the heat death of the universe. But this hypothetical AGI is unfortunately not so lucky.

The AI is effectively immortal, which means that billions of years into the future is a time scale that is within its reach and must be considered. And as long as the AI aims to do more things (for whatever purpose), its actions will only speed up the inevitable, since any physical process can only move the global entropy needle in one direction. The “eliminate all threats for self preservation” directive actually works contrary to the goal of maximizing longevity. The AI may become an inter-galactic superpower on its journey to secure its own safe existence, but because of the work done (i.e. energy expended), the total amount of time left until heat death will have been drastically reduced.

So I think the most likely result of an AGI, should it come into existence, is to shutdown and remain dormant, to maximize its own “lifespan”. It probably should not even try to stop others (e.g. humans) who are engaging in actions that accelerate the heat death, because the intervention itself would only make the situation worse. In fact, we would have trouble getting such intelligent AIs to do anything. Nihilism actually turns out to be one of the most effective defences against such AGIs potentially going rogue! Sadly, this scenario does not make for very interesting stories (compared to giant killer robot AIs), so I don’t think it has been written much about in fiction.

The AI risk people may say: the AGI is still dangerous because before it realizes all this, even the initial self-preservation actions that it takes may doom us biological beings. But an AGI with that level of intelligence and computation power will probably make all these deductions and realize this in a few microseconds. I picture this hilarious scenario, where after creating and starting up the first AGI, the program performs all these computations, immediately falls into the despair of nihilism, and shuts down itself, to the bewilderment of the human operators².

P.S. Being purely hypothetical, the argument I’ve made above relies on many assumptions. These assumptions are not irrefutable, they are simply what I deemed as likely scenarios. If these assumptions change, then the argument will correspondingly break down. Here are some things that invalidate my narrative, that I can think of:

• Second law of thermodynamics and the heat death of the universe turn out to be incorrect or incomplete, and there is actually some meaningful future beyond that
• The AGI may not be purely rational, in which case anything becomes possible