You Can’t Escape Biology

The AI computer virus is coming

Written by J. Rayner-Hilles and Trent Sullivan.

A recent poll from the AI Policy Institute found that 86% of American voters already believe AI could accidentally cause a catastrophic event. If it does, it will likely be via Malware, what in common parlance are termed computer viruses. In the next few years, the term “AI computer virus” will become as prominent in public discussion as terms like “COVID-19”, “climate change” or “war with Russia”. 

The AI computer virus is an existential threat to the modern world. When it emerges, there will be a frantic effort to try to close Pandora’s box and bolster our technological defenses. But once we reach the stage of mass panic – as we behold much of the world’s technological infrastructure shutting down or going offline – it will surely be too late to do anything. 

Along with the ability to answer general knowledge questions, compose essays, write (bad) poems, offer jokes and solve maths problems, you might be aware that ChatGPT can generate code. Consider the possibility that an AI tool like ChatGPT could be used to generate the specific code of a computer virus. This isn’t theoretical: according to AI cybersecurity firm SlashNext, a conversational AI service called “WormGPT” is already being sold illicitly. It is designed to help malicious people generate computer viruses, as well as phishing text to manipulate people. 

However, an AI computer virus would not just be a computer virus created by WormGPT: it would be one that uses WormGPT. An AI virus is defined by its ability to constantly recode itself or generate entirely novel code, such that it might modify or add to itself, or even spawn something entirely new. It is a virus that contains code generation capabilities within itself (even if that requires the internet). A traditional computer virus, on the other hand, is just a dead script. It is handwritten by the human developer who made it and is incapable of changing and adapting itself to the environment in which it spreads. For this reason the AI computer virus is something radically different: It’s alive! It can change itself, it can evolve; it can spontaneously create something new. This is why AI viruses are so dangerous; a true AI computer virus would quickly fall out of the control of its creator.

Computers are increasingly programming themselves. AI code generation has sparked a revolution in the software engineering industry. There are services like ChatGPT, such as “GitHub Copilot” and “JetBrains AI Assistant”, that have been specially trained and integrated into the applications that software developers use. Whereas all previous software was compiled from code 100% written by humans, the paradigm in the industry is now shifting. That “100%” figure is coming down; code is being written less by humans and more by AI. It is even now entirely possible for software to be written entirely by AI.

Such software isn’t generally useful in practice because AI tends to slightly misinterpret requirements and consequently writes useless if not broken code. But if the goal isn’t to produce software that is useful to humans, this doesn’t matter. The code of a virus certainly doesn’t matter to humans: it has the very simple goal of merely copy-pasting itself onto other machines. This serves only a very empty “evolutionary” purpose: survival and reproduction of the virus itself. To that end, a computer virus might just as well be entirely generated by AI, even by an indefinite process of trial and error.

Popular culture has produced many dystopian future scenarios where AI robots take over the planet and attempt to enslave or destroy the human race. In this context, when we encounter conversational artificial intelligence like ChatGPT, we naturally ask ourselves, “Is it alive?”, “Is it really intelligent?”, “Does it think?”, “Is it conscious?”, "Does it desire anything? What does it think of me? Of people?". But ChatGPT and similar large language models (LLMs) do not think or reflect on their own. They certainly don’t desire or intend to do anything other than transform prompt text into reply text when instructed by the user.

To be sure, big questions concerning the nature of intelligence, consciousness and self-awareness have been prompted by the development of ChatGPT. But further questions concerning thought, desire and intention are quite premature. Conversational AI services are trained to simply consume, chew and spit out parts of the vast corpuses of publicly available digital text they’re trained on. The “sensory existence” these models know is entirely formed of the symbolic text prompts that they are fed, and their momentary span of waking life begins with the reception of such input, and ends in the production of text output on the other end, milliseconds later. They don’t want or desire anything. They don’t think; they don’t even want to think. The AI chat technology with which we’re presently familiar is in fact just as blind and dead as the self-replicating computer malware written by humans.

ChatGPT and other services are clearly intelligent and creative, but they lack the crucial spark of biology that would bring them to life. That spark is the essential Darwinian will to survive and reproduce. Computer viruses actually do have that spark, but they are not in any way intelligent or creative, so they seem to dead to us (much like actual viruses, which most biologists agree don’t constitute living organisms). But if you combine the two phenomena – a generative AI and a computer virus – you open the door to a whole new realm of possibilities. A computer virus can give the spark of “life” to AI technology, which makes it possible that something like ChatGPT actually could have “desires” and “intentions” – that it could always be “thinking” about something with an end goal in mind.

This new creature, the “living” AI virus, is very bad news. Indeed, it will skewer the technological optimism of our age. Search the works of writers like Ray Kurzweil, and you will find hope in “event-horizon” technologies such as Artificial Intelligence: a hope to transcend biological existence. Supposedly by the end of this century man will overcome all his limitations: biotechnology will first make us immortal; then nanobots will give us godlike powers. Artificial intelligence and quantum computing will give birth to virtual realities vastly richer than the physical world we presently inhabit, and we shall all digitally upload our very consciousness to that virtual world and live in everlasting bliss.

Of course, most people aren’t as optimistic as the transhumanist futurists. But all the same, the driving force of the modern age is clearly faith in technology. The looming prospect of the AI computer virus will surely sound the death knell for the pervasive societal belief in technological progress.

Progress is felt – consciously or unconsciously – to be man’s hope of transcending his biology. Progress even includes getting rid of those controversial old-fangled biological categories like sex, race, tribe and even family – through a mixture of technology, government control and “enlightened” progressive education. As the general public becomes aware of it, and gradually comes to fear it, the AI computer virus will represent “The Great Loss of Faith” in one of the main principles of progress: technology.

It will prompt a great turning away from notions of technological transcendence, and the acceptance of biology as an inherent and immutable property of the universe itself. And again: biology means Darwinism – the gritty reality that life consists in an endless game of survival and reproduction, and that it is not possible for material existence to move beyond that worldly affair.

With respect to man’s pride in having conquered nature through technology, the AI computer virus is the very thing that should not be: our own technology coming alive and betraying us in the most primitive way. It would be tolerable for man’s collective ego that “General AI” should surpass us, become God-like and pursue some ideal of existence greater than we could ever imagine – even if it should wipe us out in the process. What is more demoralising is the prospect that instead of “General AI” we only invent “Viral AI”, which pursues the most basic goals that all biological organisms have, nothing more than survival and reproduction.

Viral AI is the rude imposition of biology into a non-biological realm. It is the desecration of our metallic and silicon temples by selfish, violent and primitive ‘biosoftware’. The remarkable achievements of science and reason will be wrought for nothing. We could suffer the indignity and shame of being wiped out – not by a greater life form that we created but by a lower one.

J. Rayner-Hilles is a computer programmer. He co-authored The Past is a Future Country with Edward Dutton. Trent Sullivan is the pen name of a data scientist and researcher.

---

Support Aporia with a $6 monthly subscription and follow us on Twitter.