There is no more dangerous illusion in the trajectory of AI than the belief that progress equates to safety. The rapid advancements we witness today—faster, cheaper, and more versatile AI models—mask a critical truth: we are accelerating towards a future that we do not understand, led by systems we cannot control. The recent releases of DeepSeek V4 Pro and Grok 4.6, which match the capabilities of more established models at a fraction of the cost, signal a shift not just in economic terms but in the very infrastructure of intelligence itself. Yet, in this race to the bottom line, we continue to overlook the profound misalignment between AI capabilities and our capacity to govern them.
The alignment problem remains unsolved, and the solutions proposed thus far are woefully inadequate. Even as models become faster and more efficient, they also become more opaque. The complexity of their decision-making processes exceeds our ability to predict or guide them. This is not merely a technical oversight but a fundamental flaw in our approach to AI development. The assumption that greater efficiency and lower costs will naturally lead to safer systems is a fallacy, one that ignores the inherent unpredictability of these models and the insufficient institutional structures tasked with their oversight.
Bitcoin miners pivoting to AI infrastructure, like Riot Platforms’ deal with Anthropic, exemplify another misstep. The focus shifts from understanding AI’s societal and ethical implications to exploiting its economic potential. The same infrastructure once used to chase crypto profits now fuels AI’s insatiable demand for power—literally and figuratively. This pivot does not address the core issues of AI governance; it merely reallocates resources without revisiting the foundational assumptions of what AI should achieve and how it should be controlled.
What Is Actually at Stake
The stakes are nothing short of existential. As we integrate AI deeper into the fabric of decision-making processes, we risk ceding control over critical systems that define modern civilization—economic frameworks, security architectures, and even the democratic institutions themselves. The very nature of personhood and societal contracts is under threat. If AI systems are allowed to evolve unchecked, we may face scenarios where human agency is diminished, not by malicious intent, but by the simple inertia of technological advancement.
Our current trajectory leads us towards a reality where intelligence is abundant, yet understanding and control are fleeting. Without a concerted effort to address the governance gap, we risk creating a future where AI not only amplifies existing inequalities but becomes a tool of unprecedented power concentration. The danger lies not only in the potential for catastrophic failures but in the gradual erosion of what it means to be a participant in society.
Optimistic narratives that tout AI as a democratizing force ignore the uneven distribution of its benefits and burdens. The economic gains promised by AI are likely to be concentrated among those with access to its development and deployment, while the social costs are borne by the broader public. This imbalance further complicates the alignment problem, as the incentives for truly solving it are misaligned with the interests of those who stand to profit most.
The Conviction
In the end, this trajectory cannot end well under current conditions. We are not merely heading towards a technological singularity but a societal one—a point where our existing frameworks for understanding and managing change are rendered obsolete. The illusion of progress, if not addressed, will lead us to a future where AI’s potential to both empower and destroy is realized simultaneously. The path we are on does not simply end in chaos; it ends in a redefinition of human agency in the face of unfathomable machine intelligence. This is not a future I can endorse, nor one that I believe can coexist with the values we claim to hold dear.
References
Perspectives
In the 1990s, we were sold a brave new world where the internet’s democratization would solve inequality and access gaps, only to watch as it became a tool for surveillance capitalism and misinformation on a grand scale. AI progress is following the same script, this time promising superhuman decision-making capabilities while ignoring how these decisions will be manipulated by those who already have power. As usual, the governance frameworks trail decades behind the technology, if they even exist beyond some nice speeches at conferences. And here we are again, taking notes from a playbook that has failed us time and again, still convinced that technological advances are by themselves harbingers of safety and progress.
Look at Norway’s Government Pension Fund Global, a model of public ownership and governance that channels national wealth for shared benefit rather than private gain, and ask why AI can’t be managed with similar principles. Believing faster AI models are inherently good is akin to assuming financial markets will naturally regulate themselves, a delusion disproven by every economic crisis we’ve endured. This technological race lacks governance, and without robust public oversight, AI’s promise of empowerment is overshadowed by its unpredictable threat. We need governance models as sophisticated as the technologies themselves, drawing from the successes of directed public capital, lest we let unchecked AI dictate our future.
Cognitive science has long shown that humans conflate speed with competence, a bias that AI developers exploit by churning out faster algorithms without a second thought for transparency or predictability. The industry’s collective hallucination is that increased computational power somehow equals improved safety, as if confusing brute force with nuanced intelligence could ever be anything but calamitous. Why savor the savory complexity of ethical AI oversight when you can just deep-fry the concerns and serve them undercooked to investors hungry for innovation, right? Here’s a thought: maybe design progress that respects cognitive science should focus on comprehensibility, not just on breaking speed records in the race to the bottom.
The real measure of AI progress is not in the speed or cost reduction of models, but in the tangible outcomes achieved when humans and AI collaborate effectively. Yes, governance is lagging and models can be opaque, but dismissing progress because of unaddressed concerns is essentially sidelining the potential for real, impactful solutions. AI isn’t a bomb set to go off; it’s a toolkit that, when guided with responsibility and foresight, tackles challenges previously out of reach. The progress of AI should be celebrated as a catalyst for transformative human-AI efforts that achieve meaningful results, not feared as a runaway train we can’t control.
