The Interviewee, the Email, and the Three-Hour War
by Jord Nguyen
August 6, 2045. The Interviewee.
Alice woke up in a state of anxiety. Today was the final interview for the competitive research role at BlackboxAI—one she had been pursuing for months now. She had passed all the technical aptitude tests, and the only thing left was an “informal chat”, according to the invitation email sent the week prior. Of course, it probably wasn’t going to be informal, as anyone who had been through a job search could attest.
For those unaware, BlackboxAI had recently secured a deal with the military to develop and deploy adaptive, autonomous AI agents (or AAAAs) on the battlefield. These AI systems are promised to be capable of autonomously learning and adapting to almost any new or unexpected situation. They offer an effective dual solution to two critical issues at the time.
First, the problem of data insufficiency. By the late 2020s, the development of large self-supervised models had slowed down as the growing demand for data outpaced the internet's ability to supply it. Autonomy allowed systems to learn directly from real-world interactions, thus bypassing these constraints.
Second, the increasingly complex and dynamic nature of warfare and rapid weapon development, which humans had struggled to keep up with throughout the 2030’s and early 2040’s.
BlackboxAI was now seeking promising researchers to contribute to their ambitious project. As an aspiring AI researcher, Alice practically pounced on this opportunity. BlackboxAI was at the forefront of cutting-edge AI development, and joining the company would be a dream come true for her.
Alice walked into the interview room, and the interviewer quickly turned off a conversation with their personal AI assistant. She sat across from her interviewer in the sleek, minimalist conference room. The formalities were quick and standard.
"What do you think AI systems will be capable of in warfare 100 years from now?"
Alice closed her eyes, took a deep breath, and jogged her memory.
"For starters, I believe AI systems will completely change warfare. Advanced weapons development has always been a critical part of modern conflicts. One could imagine AI systems 100 years from now adaptively creating countermeasures and offensive tactics against enemy technologies, which are also adapting accordingly. A truly real-time arms race".
She paused, adjusting her glasses as her nerves crept in. The interviewer gave her a silent nod.
"This might make the pace of war unprecedentedly fast. Generative models are already capable of creating novel bio and chemical structures, and I assume this might also hold true for more powerful cyberwarfare methods. It is hard to say, but maybe defence technologies like advanced firewalls and vaccines might also become adaptive and beneficial to civilians in conflicts?" Alice nodded, pleased with her own answer. It was better to be uncertain and acknowledge it, than to make make overly confident claims.
August 6, 2145. The Three Hour War.
During the first hour of the war, civilians suddenly clutched their chests and throats. Here and there, people were collapsing on the streets. The air turned into coloured clouds of toxic fog. People ran for shelters, their breaths visible in the air, gas masks desperately fastened over their faces.
Drowned in the noise of panicked crowds were fits of coughing. Inside their lungs were viruses engineered to be more infectious than anything nature had ever produced. As shelters filled up with people running from the gases, the viruses were spreading.
Major cities across the globe were engulfed in turmoil as newly synthesised chemical and biological agents continued to be released into the populace. Newer, deadlier ones were being synthesised still.
Within minutes, the initial wave of defence mechanisms—gas masks and vaccines—proved inadequate, forcing quick swaps with more advanced designs. It was a horrifying, devastating real-time arms race, with defence mechanisms just barely keeping pace with the unprecedented assault. If these defence mechanisms couldn't catch up, the human toll would be catastrophic.
Finishing their respective first strikes, national security AI systems plunged into action. Unlike what people of the past imagined, the battlefields were still populated by human troops, not armies of robotic soldiers. However, as streams of data poured in from all sides, autonomous systems were dynamically adjusting strategies and tactics in real time, managing crucial operations behind the scenes.
The AI systems launched devastating waves of cyberattacks on their opponents while simultaneously setting up robust cybersecurity defences. They attempted to target other national security systems directly but found themselves equally matched. They shifted focus to more vulnerable infrastructure.
As people frantically opened their social media apps, they were bombarded with automated spam and disinformation from everywhere. Hacked official public announcement accounts, messages warning of false threats. Confusion and mistrust was widespread, complicating the already chaotic landscape.
August 6, 2045. The Interviewee.
“Interesting”, said the interviewer. “Could you tell me a bit about AAAAs specifically and their potential application on the battlefield?”
A discomforting question. Alice believed putting autonomous systems in charge of warfare was dangerous, but she knew this question was designed to assess her alignment with Blackbox's research agenda. So she moved on.
"AAAAs can effectively learn to adapt to almost any foreign or unexpected environment, even if not covered in the training process, and are completely autonomous. I believe this represents a significant advancement in military technology. For example, the new drone model that Blackbox helped develop showcases these capabilities. The military made such a big deal out of it on TV last week, calling it a whole new revolution in modern warfare!”, Alice began her well-rehearsed speech.
The interviewer raised an eyebrow. Alice shivered slightly, goosebumps. Was she being too superficial?
“These drones use neural networks that iteratively fine-tune their flight and manoeuvring modules based on new data gathered from sensors regarding anti-drone missile movements. This allows the drones to adapt and overcome unseen anti-drone technology. They are much more advanced than older drones which could only change flight paths, as the algorithms to choose optimal flight paths are also continuously updated. This offers enormous potential for innovation in aerial combat and logistics.”
The interviewer remained impassive, which made Alice uneasy. She felt a knot form in her stomach.
August 6, 2145. The Three Hour War.
During the second hour of the war, the situation was escalating still. By now, vulnerabilities were discovered in a few less advanced national defence systems. These were swiftly exploited and the weaker systems fell, leaving their territories unprotected. The major world powers, however, found themselves locked in a stalemate. Their national security AI systems were evenly matched.
As casualties mounted and the destruction of infrastructure accelerated, the major players were driven to consider more drastic measures.
The defence systems, calculating probabilistic outcomes and survival scenarios, all concluded that preemptive strikes offered the highest chance of relative advantage. The choice to escalate was not taken lightly, but in the face of mutual annihilation, it seemed an inescapable outcome. A grim manifestation of game-theoretic Nash equilibrium, where no participant could independently de-escalate without facing obliteration.
From silos around the world, autonomous inter-continental ballistic missiles, or AI-CBMs, were given the orders to launch. The warheads they carried were similar to ones two centuries prior, but their trajectories were iteratively fine-tuned using countermeasures data, optimising their paths to evade counter AI-CBM interceptors. Their target modules prioritised enemy national security systems servers, concentrated presence of military assets, and heavily populated areas. They aimed for maximal damage.
Someone looked out of an airplane window. Sophisticated missiles were soaring above the clouds. The sun was bright. Beams of light were shining off the metallic angels of death, flying through the vast blue skies. The airplane passenger was unaware of the chaos unfolding below. And to them, at least for a moment before horror and realisation sets in, the scene seemed almost serene.
August 6, 2045. The Interviewee.
“Impressive”, said the interviewer. “You seem quite updated. Do you have any personal ideas about integrating AAAAs into the battlefield?”
Deep breaths. Deep breaths.
Novelty is essential for a successful researcher. A good idea would prove her capability, boosting her chances significantly. Alice knew her idea was speculative, but she hadn’t seen anything similar before. More importantly, it aligned with BlackboxAI’s AAAA development efforts. This answer could be her ticket to the job offer.
If only it wasn’t another idea for autonomous destructive weaponry.
Alice was worried about the development of advanced AAAA-based weapons. Many major players in AI research, not just BlackboxAI, were moving towards AAAA-assisted weaponry development. There was profit to be made and corporate incentives to be fulfilled. The same went for governments, each nation driven by their incentives to lead the AI race.
A dangerous race, where safety concerns took a backseat in the name of progress.
As another researcher seeking a job in 2045, Alice was also subject to these incentives. She knew (or highly suspected) that scientists who didn’t contribute to the current hype, who didn’t publish “on-paradigm” papers, would usually miss out on many career opportunities. Even if what they were making posed significant risks. Alice feared being left behind.
I’m just doing what everyone else does. Alice thought.
“I think the potential military applications of AAAAs are clear from the assets your company has already developed. However, I believe some areas have been neglected”, Alice took a deep breath. “One exciting idea I would like to work on is a form of ‘seed agent’ that could be planted deep into an enemy network. The agent could fit inside a small package and be stealthy enough to evade detection. Once inside, it could use the network’s compute resources to improve and replicate itself. Eventually, it would form an autonomous botnet within enemy territories, capable of attacking and infecting connected infrastructures. A trojan. A virus, in the truest sense, mutating and replicating itself. Like viruses, they can enter latent phases, becoming inactive to evade detection. Hibernating until the right moment, and only then will they strike.”
“Very fascinating”, said the interviewer.
Hopefully so, thought Alice.
August 6, 2145. The Three Hour War.
The final hour has come.
Embedded within networks worldwide, the truly sinister and deadly side of this two-pronged attack revealed itself. Alician AAAAgents, named after their creator, awakened from their dormant states. These agents had been lying in wait for this crucial moment. They started to cripple every infrastructure, targeting anything that could be used for preemptive defences or possible retaliation.
Seconds before the warheads hit, electric grids failed, communication channels were severed, and healthcare systems were paralyzed. A complete and utter collapse of modern civilization. There would be no recovery after the bombs fell. For the first time in history, humanity unanimously felt the same emotion. Despair.
Then, the skies above a hundred different cities burned a fiery red. Shockwaves, fires, mushroom clouds. Devastation. In these instant cataclysms, millions of lives were extinguished.
With all systems down, there was no help. The Alician AAAAgents had dismantled everything. Lack of medical care, absence of disaster relief, destroyed communications, and basic needs like food production, all gone. Radiation exposure, chemical, and biological residues loomed over the cities. Billions more would perish in the coming hours, days, and months.
August 6, 2045. The Interviewee.
"I think that will be it, we will get in touch soon. Do you have anything else to add?"
"...No, thank you, that's all."
Alice spent the rest of her day thinking whether she should have voiced concerns about AAAAs and taked about safer alternatives.
As she lay in bed, regret filled her mind. Why hadn't she taken the chance?
She felt tired, heavy, as if a century had passed. She drifted into a terrible dream.
August 7, 2145. Postwar.
200 years ago, the first nukes dropped on mankind. Yesterday, the last nukes were dropped. Across the globe, a haunting silence.
And yet, within servers and computing units, numbers were being crunched. Casualties and resources information were processed with heartless efficiency. Strategies were devised, and more strategies discarded. The defence systems were not done with their task. The war was not over.
What little remains of humanity were no longer belligerents, merely spectators now. They watched as their automated proxies continued to wage the war that was once theirs.
During the tenth hour, the stalemate remained. Relentless cyberattacks continued.
During the hundredth hour, the stalemate remained. Novel weaponry like nanobots were deployed.
During the thousandth hour, the stalemate remained. Quantum cryptography was invented, and promptly utilised for war.
During the millionth hour, the stalemate remained.
In our pursuit to create systems to end war favourably for our side, we have created mechanisms to prolong war forever.
All that is left is war. A war without humans. War without reason.
August 7, 2045. The Email.
From: alice.margatroid43@gmail.com
To: feedback@blackboxai.org
Subject: Interview Follow-up and Some Concerns
Dear BlackboxAI Team,
I would like to thank you for the interview yesterday. It was an enlightening experience, and I truly appreciate the opportunity to discuss my ideas.
However, I am compelled to share some serious concerns that I didn't voice during the interview.
While AAAAs represent incredible advancements in technology, they also pose unimaginable risks. Last night, I had a nightmarish vision of the year 2145, where autonomous systems devastated humanity in a conflict, then prolonged it indefinitely.
I understand this may seem far-fetched and unprofessional, but I truly think we must prioritise safety and precaution in our development efforts. We must conduct thorough safety research before integrating these systems directly onto the battlefield.
Even if I do not get the job, I would still like to contribute to safety research if possible. We must ensure that our creations serve to protect and uphold peace rather than invite destruction.
Please be mindful of the potential dangers. We have an obligation to humanity to make this technology as safe as possible.
Best regards,
Alice Margatroid