The Semantic Vacuum of the Slowdown Debate
In recent months, the tech industry has become obsessed with a single, vaguely defined concept: the "AI slowdown." High-profile figures and former government advisors have floated the idea of slowing the pace of development to prevent existential risks or market bubbles. Yet, when pressed, the proponents of this deceleration offer little more than abstract frameworks. Some suggest a three-point plan involving independent monitoring, industry-wide regulation, and global governance. Others simply worry about a potential burst bubble in the venture capital sector. Despite these discussions, no one has provided a substantive definition of what a slowdown actually looks like in practice. Ed Zitron, CEO of EZ Primary Research, has pointed out that the industry is currently running headlong into a wall of massive spending without a clear path to revenue, yet the discourse remains stuck in a loop of corporate speculation.
This lack of clarity is not just a technical oversight; it is a structural failure. When we talk about "slowing down," the conversation almost exclusively orbits around two poles: the fear of a rogue superintelligence (the existential risk) and the fear of a massive financial crash (the market risk). These are quintessentially Western, anthropocentric concerns. They focus on the survival of the human species or the stability of Wall Street, while ignoring the very tangible, immediate wreckage being caused by the current acceleration.
The Material Cost of Silicon Dreams
To understand what a slowdown could actually achieve, we must look away from the server farms in Northern Virginia and toward the copper mines in the Democratic Republic of the Congo and the water-stressed regions of the American Southwest. The current AI arms race is a physical process. Training a single large language model requires millions of gallons of fresh water for cooling and an enormous amount of electricity. It also demands a constant stream of rare earth minerals, cobalt, and lithium.
For the Global South, the accelerating pace of AI development is not a theoretical debate about safety; it is a direct extraction of local resources. Massive amounts of energy are diverted to power the data centers of a few trillion-dollar companies, often in regions where local populations still face energy poverty. Water used to cool a GPU cluster in a drought-stricken area is water that isn't going to crops or local communities. An "AI slowdown" that focuses only on slowing down code development does nothing to address this resource diversion. A true slowdown would need to be a deceleration of extraction—a pause on the relentless hunger for more compute power that drives environmental degradation in vulnerable ecosystems.
Labor and the Myth of Autonomous Intelligence
There is a pervasive myth in Silicon Valley that AI is a purely digital phenomenon, a ghost in the machine that operates independently of human labor. This is a lie. The polished interfaces of modern AI models rely on a massive, invisible workforce of data labelers and content moderators, many of whom are located in the Global South. These workers spend hours performing the repetitive, often traumatic task of tagging images or filtering graphic content to make models "safe" for Western users.
The current acceleration forces these workers to work faster, for less, under increasing pressure to keep up with the training cycles of new models. When proponents of an AI slowdown talk about "," they rarely mention the working conditions of those in Kenya, the Philippines, or Venezuela. If a slowdown is merely about regulatory oversight for software, it remains an elite concern. A meaningful slowdown would address the labor exploitation that makes the current pace of development possible. It would treat the people behind the data as central to the conversation, rather than as invisible infrastructure.
The Trap of Centralized Regulation
There is a danger that the current push for global regulation will actually entrench the power of the current tech giants. This is a phenomenon often seen in highly regulated industries: the largest players have the capital to navigate complex legal frameworks, while smaller competitors are crushed by the compliance costs. If we implement heavy-handed, top-down regulations under the guise of a "slowdown," we might inadvertently create a moat around the existing incumbents.
A slowdown that favors only the giants will stifle grassroots, community-driven innovation. True technological sovereignty—the ability for communities to build tools that serve their specific local needs rather than global advertising models—requires more than just "slowing down" the big players. It requires redistributing the power and the resources they currently hoard. If a slowdown doesn't facilitate a shift toward decentralized, open-source, and locally-governed technology, it is just another tool for corporate consolidation.
Redefining Safety: From Existential to Tangible Risks
The term "existential risk" is often used to hijack the attention of policymakers. It shifts the focus from the harms we can see today toward a hypothetical catastrophe in the future. While we should prepare for long-term risks, prioritizing them at the expense of current realities is a political choice. It allows companies to evade responsibility for the algorithmic bias, data theft, and environmental costs that are happening right now.
We need a new definition of a slowdown. Instead of measuring the speed of
