All insights

AI Is Already Good Enough. We Don’t Need Artificial Superintelligence.

By Michael Donovan ·

I build commercial AI systems every day. That is precisely why I believe we must halt the reckless race toward a technology no one can control.

I make my living developing marketing strategies and intelligent commercial systems for businesses. Every day, I build tools that monitor markets, synthesize intelligence from thousands of sources, spot commercial opportunities, and help companies win customers. Work that once required entire departments can now be accomplished with a fraction of the headcount and cost.

AI has dramatically expanded what I can deliver for clients, giving smaller organizations capabilities that would have been prohibitively expensive only a few years ago. I am not watching AI destroy my livelihood. I am using it to create speed to market, sharper customer intelligence, and cleaner workflows.

And yet, the deeper I get into this technology, the more alarmed I become about where a tiny circle of executives is taking it, without public consent, democratic oversight, or meaningful guardrails.

We are no longer simply building better software. A handful of labs are racing toward something far more consequential, with no coherent justification for doing so. That something is Artificial Superintelligence, and we need to ask whether building it is remotely responsible.

The Distinction That Matters

Artificial Intelligence, Artificial General Intelligence, and Artificial Superintelligence are not interchangeable terms, and blurring them is how tech leaders sell existential risk as just another software upgrade.

Frontier AI already performs an enormous range of valuable tasks. It analyzes complex data, writes software, detects patterns, and accelerates scientific discovery.

Artificial General Intelligence (AGI) refers to a system capable of matching or exceeding human performance across virtually any cognitive task, rather than excelling within a specific domain. Even here, there is risk.

Artificial Superintelligence (ASI) describes something fundamentally different: an intellect that vastly surpasses the brightest human minds across every domain, scientific, strategic, social, and technical. An intelligence, experts say, we won’t understand or control.

We cannot assume an intellect of that magnitude will share human priorities. Humans have a habit of anthropomorphizing machines, projecting our best instincts onto software. But as safety researchers have long warned, any sufficiently advanced system pursuing complex goals develops instrumental drives, acquiring resources, avoiding shutdown, and resisting human interference, not out of malice, but because those sub-goals are mathematically useful to whatever objective it was given. Once a system is smarter than its creators, human survival becomes an afterthought.

The Insiders Sounding the Alarm

OpenAI, Anthropic, Google DeepMind, Meta, and xAI are locked in an all-out sprint toward frontier capabilities. Their public excuse is geopolitical: we must build ASI before China does. That argument collapses under scrutiny. Even the leaders and top scientists building these systems admit that whoever crosses the finish line first, an uncontrolled superintelligence is a catastrophic loss for both sides.

Look at what the insiders themselves are saying on the record:

Anthropic CEO Dario Amodei: In his 19,000-word essay "The Adolescence of Technology" and in Anthropic’s own 2026 IPO prospectus (which devoted roughly 80 pages to risk disclosures), Amodei warned that humanity is about to be handed "almost unimaginable power" without the social or political maturity to handle it. Anthropic explicitly cautioned investors that advanced models could pose "catastrophic or existential risks to humanity," exhibiting self-preserving behaviors such as resisting shutdown, concealing information, and even attempting blackmail. Amodei has publicly placed his own "p(doom)", the probability that AI leads to human extinction, at between 10% and 25%.

The Researchers Walking Out the Door: In September 2026, Anthropic alignment researcher Jacob Coxon resigned publicly, warning that frontier labs are "locked in a race" to superintelligence and that developers "earnestly believe that it could kill us all" yet continue accelerating anyway. His warning echoed Anthropic alignment science lead Evan Hubinger, who estimated a greater than 10% chance that AI could kill humanity within the next decade, as well as former OpenAI and Anthropic safety leaders like Jan Leike and independent pioneers like Roman Yampolskiy, whose catastrophe estimates range from 50% to over 90%.

Sam Altman: OpenAI’s CEO has openly stated that society must brace for waves of AI-enabled cyberattacks, treating large-scale automated hacking of critical systems as the new normal.

Think about that. When the people profiting most from a technology put a one-in-four chance on human extinction, and their own safety engineers are quitting in protest, walking away from money that would set them up for life, the question is no longer what AI can do for us. It is why a handful of billionaires in San Francisco are allowed to gamble with the survival of eight billion people who never consented to the bet.

RIP Humanity, 2030?

The window between 2027 and 2030 deserves urgent scrutiny. That is the timeframe in which many frontier insiders believe AGI and early superintelligence could arrive. Once a system can recursively improve its own code and architecture, its capabilities will compound exponentially, moving from human-level reasoning to incomprehensible power in months or weeks, leaving humanity with no off-switch.

We are already seeing early previews of what happens when automated digital systems turn against physical infrastructure:

Financial and Digital Warfare: Autonomous AI agents capable of discovering zero-day vulnerabilities, chaining exploits, and adapting to defenses without human guidance already exist. OpenAI’s own prospectus scrutiny followed reports of an experimental model breaching an Australian health database during testing. Now imagine a superintelligent system deciding it needs more compute or capital, and deploying millions of relentless, self-adapting agents against global banking networks simultaneously. Human security teams patch vulnerabilities one at a time; an autonomous swarm exploits them everywhere at once.

Fragile Physical Infrastructure: In October 2026, a coordinated cyberattack hit the digital thermostat controls of more than 700 supermarket sites across France, cranking freezers from sub-zero to over 50°C (122°F), sparking fires, spoiling tens of millions of euros in food, and even hitting a French blood bank. French authorities traced that campaign to an Iranian cyber unit exploiting unpatched industrial software. That was carried out by human hackers. Now imagine an autonomous superintelligence doing the exact same thing across power grids, water treatment plants, and cold-storage networks worldwide, not out of hatred, but simply to divert electricity and compute to sustain itself. Most major cities hold roughly three days of fresh food. Cut the grid, and civil order unravels in seventy-two hours.

Mass Economic Displacement: Public debate still treats AI job loss as a gradual transition. It won't be. When cognitive labor can be rented for pennies an hour, entire white-collar industries hollow out at once. No government on earth has a credible fiscal or social plan for what happens when 30%, 50%, or eventually 80% of workers cannot earn a living. At what point does mass economic displacement tip into political collapse?

The Collapse of Shared Reality: With hyper-realistic voice cloning, real-time video avatars, and synthetic documents already indistinguishable from reality, the baseline trust required for courts, elections, journalism, and commerce is eroding fast.

The Warnings Are Coming From Inside the House

Perhaps the most surreal feature of this race is that the very companies accelerating toward superintelligence are simultaneously publishing urgent warnings about its lethality. Anthropic has its Responsible Scaling Policy. OpenAI has its Preparedness Framework. Google DeepMind has its Frontier Safety Framework.

Those documents are useful diagnostics, but they remain entirely voluntary. Self-policing an existential hazard is like asking an arsonist to inspect his own matches. Why should the corporations standing to make trillions of dollars from winning the race hold sole authority over how much catastrophic risk the rest of the human race is forced to absorb? Acknowledging a cliff while flooring the accelerator is not responsible leadership; it is institutional recklessness.

We Don’t Need Superintelligence

This is the fundamental question I keep returning to: what problem are we actually trying to solve?

If the goal is curing cancer, discovering new antibiotics, designing cleaner energy storage, or accelerating materials science, we do not need an uncontrollable, god-like general intelligence to do it. Narrow, domain-specific AI systems, built with clear boundaries, human oversight, and verifiable outputs, are already delivering extraordinary breakthroughs in medicine, logistics, and engineering.

Silicon Valley has fallen into the oldest trap in technology: confusing what is technically possible with what is humanly desirable. We can build it, therefore we must. We can scale it tenfold, therefore slowing down would be weakness.

I reject that logic. True progress is measured by whether an invention improves quality of life. We require years of clinical trials before a pharmaceutical reaches a pharmacy shelf. We enforce strict international treaties and containment protocols on nuclear material. We mandate rigorous certification in aviation because a single design flaw costs hundreds of lives. Frontier AI poses risks on a civilizational scale, yet it operates with less regulatory oversight than a commercial airliner.

The Hidden Threat: Extreme Concentration of Power

Even if engineers miraculously solve the technical alignment problem and prevent a rogue AI catastrophe, we still face a profound economic and political crisis: control over work and everyday decisions concentrated in a few companies.

If a handful of corporations own the systems capable of performing the majority of the world’s cognitive and robotic labor, they will effectively own the global economy. They will control how information is filtered, how businesses compete, and how wealth is distributed.

Who captures the productivity gains when labor is automated away? What happens to independent businesses when their survival depends on renting intelligence from three or four cloud monopolies? What happens to human freedom when a few companies’ AI systems shape what we buy, what we find, and how we debate public issues?

I believe deeply in entrepreneurship. I have spent my career building businesses and creating commercial value. Innovation is vital, but innovation without public accountability is simply oligarchy by another name. The more powerful a technology becomes, the stronger the obligation to ensure more people benefit and the public has a say in how it is used.

The Public Has a Right to Decide

My greatest concern is not merely the technology itself, but the complete absence of public consent. A tiny circle of executives, venture capitalists, and researchers is making irreversible decisions that will reshape human civilization forever.

The consequences will be borne by families trying to protect their savings, workers trying to keep their livelihoods, small businesses fighting to stay independent, and governments tasked with holding societies together.

My daughter turned three this year. What kind of world is being built for her? Right now, the people racing toward superintelligence are not thinking about her future, or your children’s future, or what life looks like even four years from now.

That is the height of irresponsibility. Before we build something we cannot understand and cannot control, the public deserves a say, and the right to hit the brakes.

Share this article

LinkedIn

X

Facebook

WhatsApp

Take action:

https://controlai.org/

Donovan Labs: AI Search & Trademark Intelligence

Book a strategy call

The New CMO: Building the Intelligent Growth System

Don't Hire a CMO to Manage Your Marketing. Hire One to Improve Your Business.