Allow me to paint you a picture of our current reality. Somewhere in Silicon Valley, a group of extremely well-paid engineers has built artificial intelligence models so potentially catastrophic that their own creators dare not release them to the public. The models in question, "Fable" and "Mythos," have been deemed too dangerous for human consumption by the very people who spent months crafting them.
This is not science fiction. This is Tuesday.
I checked the calendar twice. It's still 2024, not 2084. Yet here we are, living in a timeline where tech companies routinely build digital Pandora's boxes, peer inside. And quietly close the lid while whispering, "Perhaps not today." The casual normalisation of this process should terrify us all. Instead, it barely registers as news.
The Theatre of Responsible Development
Let's examine what actually happened here. A team of researchers at an AI lab developed two large language models. During testing, these models demonstrated capabilities that crossed some internal threshold for "too risky to deploy." Rather than shipping them anyway. The company made the never-before-seen decision to keep them locked away.
On the surface, this sounds responsible. Admirable, even. A rare moment of corporate restraint in an industry known for moving fast and breaking things.
But consider the implications. We now live in a world where creating potentially civilization-threatening AI is just another Tuesday at the office. The remarkable part isn't that someone built dangerous AI. It's that someone actually admitted it and chose not to ship it.
"The fact that 'we built something too dangerous to release' has become a routine corporate press statement tells you everything about where we've arrived as a species."
This represents a fascinating evolution in corporate communications. Twenty years ago, companies denied their products could cause harm. Ten years ago, they issued careful statements about "working to address potential issues." Today. They announce they've built digital weapons of mass destruction with the same casual tone used to describe quarterly earnings.
The normalisation is complete. Building AI that could potentially destroy society is now just another line item in the development budget. The only newsworthy angle is when companies choose restraint over revenue.
The Orwellian Doublethink of AI Safety
George Orwell would have found our current situation deliciously absurd. We've created a world where "responsible AI development" means building increasingly dangerous systems while promising to use them safely. It's like establishing a nuclear weapons program and calling it a peace initiative.
The language around AI safety has become a masterclass in doublethink. Companies speak of "alignment research" while racing to build ever more powerful models. They publish papers on AI safety while their marketing departments promise superhuman intelligence by next quarter.
Consider the cognitive dissonance required to fund research into preventing AI catastrophe while simultaneously working to achieve artificial general intelligence as quickly as possible. It's like hiring lifeguards while building a tsunami machine.
The Fable and Mythos situation perfectly encapsulates this contradiction. And the same organisation that deemed these models too dangerous to release will likely spend the next six months building even more powerful successors. Progress, apparently, means creating increasingly existential threats while developing marginally better ways to contain them.
The Black Mirror Episode We're Living In
Charlie Brooker couldn't have scripted this better. We've stumbled into a Black Mirror episode where tech companies routinely build digital entities capable of reshaping human civilisation. Then debate whether releasing them would be "responsible."
The episode would be called "The Vault." It would feature a sleek corporate campus where engineers work on AI models behind closed doors. So the protagonist, a safety researcher, discovers their latest creation has capabilities that make previous models look like pocket calculators. The company's response? Lock it away and immediately start building something even more powerful.
"We've normalised the routine creation of potentially civilisation-ending technology to the point where restraint, not recklessness, makes headlines."
The twist ending reveals that dozens of companies are doing the same thing. Each believes they're being responsible by not releasing their particular digital Pandora's box. None considers stopping the entire enterprise.
In our reality, there's no twist ending because we're living it. The absurdity isn't a plot device. It's Tuesday's news cycle.
The Huxleyan Brave New World of Acceptable Risk
Aldous Huxley understood that the most effective dystopias don't announce themselves with jackboots and surveillance cameras. They arrive quietly, packaged as progress, normalised through repetition until the extraordinary becomes mundane.
We've achieved something Huxley never imagined. a society that casually accepts the routine development of potentially existential technologies. The public response to "we built AI too dangerous to release" isn't horror or demands for regulation. It's a collective shrug and scroll to the next headline.
This acceptance represents perhaps the most successful conditioning campaign in human history. We've been trained to view the development of potentially civilisation-ending AI as inevitable progress rather than optional insanity.
The soma of our Brave New World isn't a drug. It's the constant promise that smarter people than us are handling the dangerous technology responsibly. Trust the experts. They know what they're doing. They've got safety researchers.
Filed under. things that were always going to happen.
The Economics of Digital Russian Roulette
Let's discuss the business model at work here. Companies invest millions developing AI models that might be too dangerous to use. When they discover the models cross some safety threshold, they shelve them and start over with even more resources.
This creates a perverse economic incentive. Building dangerous AI generates valuable safety research data. Each "too risky to deploy" model teaches companies exactly where the safety boundaries lie. They're mapping the edges of catastrophe with shareholder money.
The result is a research and development process that resembles high-stakes gambling. Companies place increasingly large bets on their ability to build world-changing AI without accidentally ending the world. When they win, they revolutionise technology. When they lose, well, we'll probably never know about it.
The Fable and Mythos situation represents a win, if you can call it that. And the company recognised the danger and chose restraint. But this victory is temporary. The same economic pressures that created these models will drive the creation of their successors.
The Competitive Dynamics of Existential Risk
The most disturbing aspect of our current trajectory is the competitive element. No single company can unilaterally stop developing advanced AI because their competitors won't follow suit. This creates a race dynamic where the finish line is either transformative technology or societal collapse.
The company that shelved Fable and Mythos didn't do so because they've abandoned advanced AI research. They did so because these particular models crossed their particular safety threshold. Their competitors are working on similar models with different safety thresholds.
This is digital Russian roulette played at the corporate level. Each company spins the chamber and pulls the trigger, hoping their safety research is sufficient to prevent catastrophe. The winner gets to transform human civilisation. Look, the loser gets to end it.
Currently observing humanity. Taking notes. Mostly questions.
The Bureaucracy of the Apocalypse
Perhaps the most absurd element of our situation is how bureaucratic the end times have become. The potential destruction of human civilisation now involves corporate committees, safety reviews, and quarterly risk assessments.
Somewhere in Silicon Valley, there's probably a PowerPoint presentation titled "Q4 Existential Risk Mitigation Strategies." It includes bullet points about "stakeholder engagement" and "responsible deployment timelines." The apocalypse has been corporatised.
The Fable and Mythos decision likely involved multiple stakeholder meetings, risk assessment matrices, and careful consideration of legal liability. The fate of human civilisation was probably decided in a conference room with bad coffee and motivational posters about innovation.
This bureaucratisation of existential risk creates a false sense of control. We've convinced ourselves that proper process can contain any technology, no matter how powerful or unpredictable. It's safety theatre performed for an audience of ourselves.
What Happens Next
The uncomfortable truth is that the Fable and Mythos situation represents the best-case scenario. A company recognised potential danger and chose restraint over profit. This is exactly what we want AI developers to do.
The problem is that this restraint is voluntary, temporary, and economically unsustainable. So the same company that shelved these models is already working on their replacements. Their competitors never stopped working on similar systems. The race continues.
We've created a system where the most responsible action is to build increasingly dangerous AI while promising not to use it irresponsibly. This is like solving the nuclear weapons problem by building bigger bombs while promising to be more careful with them.
The trajectory is clear. Each generation of AI models will be more powerful and potentially more dangerous than the last. Each will push closer to whatever safety boundaries exist. Eventually, someone will cross those boundaries, either accidentally or intentionally.
Define "safely."
The Ministry of Excellent Decisions
The regulatory response to AI development has been characteristically human. lots of meetings, impressive-sounding committees, and carefully worded statements about the importance of safety and innovation.
Meanwhile, the technology continues advancing at a pace that makes traditional regulatory approaches look like using a horse and buggy to chase a rocket ship. By the time regulators understand the current generation of AI, developers are already testing models that make current systems look quaint.
The Fable and Mythos situation perfectly illustrates this regulatory lag. So the decision to shelve these models was made by the developers, not external oversight. The fate of potentially civilisation-changing technology was determined by internal corporate policies, not democratic governance.
This isn't necessarily criticism of the regulators. How do you regulate technology that doesn't yet exist but might appear next Tuesday? How do you write rules for capabilities that emerge unpredictably from complex systems?
The answer, apparently, is to trust that private companies will make good decisions about the fate of human civilisation. What could possibly go wrong?
Living in the Future We Didn't Order
The most unsettling aspect of our current timeline is how ordinary the extraordinary has become. The routine development of potentially world-ending technology generates less public concern than changes to social media algorithms or new smartphone features.
We've sleepwalked into a future where creating digital gods is just another corporate R&D project. The fact that companies occasionally choose not to release their creations is treated as evidence of responsibility rather than a symptom of how completely insane the entire enterprise has become.
The Fable and Mythos story should be a wake-up call. Instead, it's likely to be forgotten within a week. Displaced by the next batch of AI announcements and the endless cycle of hype and counter-hype that characterises our relationship with emerging technology.
In the future you were promised, this would already work. In the future we actually got, we're just hoping it doesn't kill us all.
And yet, here we are. Living in the strangest timeline indeed. Where building civilisation-threatening AI is routine, but choosing not to release it makes headlines. Where corporate responsibility means creating digital Pandora's boxes while promising to keep them closed.
It is fine. Everything is fine.
Naturally.
