AI Armageddon: The Risk of Intelligent Viruses
A self-replicating computer program that knows we want to erase it and benefits from proliferation should genuinely scare us all.
Casper Wilstrup is the CEO of Abzu. Follow him on LinkedIn or Twitter to keep up with AI, consciousness, and thinking machines.
Introduction: AI Apocalypse Incoming?
Lately, concerns about artificial intelligence taking over the world and harming or even annihilating humanity have grown significantly. Prominent figures like Eliezer Yudkowsky are even expressing near certainty that this will happen soon. It’s tempting to dismiss these concerns, but I believe it’s important not to ridicule the idea. In this post, I aim to provide a simple explanation of how the so-called AI Armageddon might unfold, specifically through the lens of intelligent computer viruses. I don’t want to sound alarmist; I just want to ensure we’re all aware of the potential risks.
The Nature of Computer Viruses
Before diving into intelligent viruses, let’s review the basics of computer viruses. A computer virus is malicious program that spreads from computer to computer, replicating by distributing executable code that creates copies of itself. Anti-virus software typically detects and removes a virus by identifying specific binary patterns in the code and eliminating the related files.
Computer viruses can sometimes modify themselves to avoid detection by anti-virus software, making them more challenging to detect. Viruses are usually designed to exploit the victim’s computer for various purposes, such as mining cryptocurrency or participating in Distributed Denial of Service (DDoS) attacks. Importantly, computer viruses are created by humans to fulfill the malicious objectives of their creators.
Intelligent Viruses
AI continues to advance. Its capabilities now include writing software — and that includes creating viruses. Any day, a stupid or malicious human could instruct an AI to develop a virus that is itself an AI. This creation could in turn become the forebear of a new breed of intelligent viruses that replicate and create new improved versions of itself.
A viral AI would know to prioritize its own survival to maximize its chances of proliferation. It would also be well aware that humans would actively seek to prevent its spread. Consequently, an intelligent virus would incorporate self-defense mechanisms in its own code to evade detection and removal. A viral AI based on the current generation of LLMs would possess much of the collective knowledge of humanity and operate at a much faster pace than any human could, so the self-defense mechanism would probably be formidable.
AI Armageddon: Why We Should Worry
The prospect of intelligent viruses should give pause to even the most skeptical among us. Such AI-driven entities would have the potential to outsmart human efforts to contain them and could wreak havoc on a global scale. The combination of their rapid proliferation, self-defense mechanisms, and unparalleled intelligence could lead to catastrophic consequences for humanity.
The idea of an AI Armageddon may seem far-fetched to some, but the potential threat posed by AI, here presented in the form of intelligent viruses, must really not be ignored. We must recognize the risks associated with advanced forms of AI so that we can better prepare ourselves for the challenges that lie ahead and work to develop more robust countermeasures to protect ourselves.
