The Opportunity Cost of “Pacing the Frontier”

2026-09-18

The discussion that started with Jacob Coxon’s post and what followed Dario Amodei’s “We must pace the frontier” essay got me thinking.

I am personally against this mini-movement of slowing down AI development, for reasons far less practical than the obvious “if you want to stop, then stop. You don’t need permission”. I am against it for fundamental reasons: human progress and civilisational development.

Right now, the progress of artificial intelligence, along with robotics, crypto, and quantum computing, is at the absolute forefront of advancement. These technologies are bringing a radically (positively) different, more advanced world.

What does “pace” mean here? It means slowing down. Until when? Until an unspecified point in the future, when an unspecified threshold is met. I guess, until a group of objectively very smart people and experts in their fields decide to accelerate again. But these people don't hold a god-given truth; they are prone to mistakes like everyone else, and their worldview of what is right or wrong, good or bad, can differ greatly from the majority. Being an expert in a field, even as complex as artificial intelligence, does not make you see the future. Primarily because the future is the sum of many different events that happen in fields they are not experts in.

They call for slowing down because they see a theoretical threat to humanity. But in reality, they have failed to explain how humanity is threatened beyond speculative risks. For example, they talk about the possibility of a bioweapon. But that would require an unsupervised, advanced, fully automated laboratory in software and hardware. I am removing the human factor from the scenario because teams already exist that can operate sophisticated laboratories and possess enormous amounts of the required knowledge, yet we have not seen a civilisation-threatening engineered biological event. To be honest, the AI models consumers and enterprises use every day are far from capable of achieving such a feat. Maybe the AI companies have developed models 10 times more advanced than the ones they have commercialised? I doubt it, but this is what their lack of transparency brings.

Their position is also fundamentally contradictory. They champion AI's potential to cure diseases and spur economic growth; so they essentially demand a slowdown that would inevitably hamper critical breakthroughs like drug discovery. They are effectively asking to delay progress on solving real-world problems over theoretical risks that currently have no signs of happening, other than the OpenAI–Hugging Face incident, which, when you read the technical account, makes you open-weight pilled and you question whether OpenAI’s DevOps know what internet isolation means. That very incident and the refusal of so-called frontier models to help HF with defensive work is itself an argument against what this mini-movement believes are good guardrails.

We as humans have survived because we advanced in science and technology. So essentially, these people are asking for the human race to stop what has kept it alive until an unspecified time in the future. This framing alone should make every pro-human person against this movement. There have always been dangers created by technological advancement. Each one has theoretically threatened humanity. But this is the cost of advancement. The cost of progress is the corresponding responsibility to govern use the new invention for good. But the real threat comes from stopping humanity from doing what it does best. Advance.

I say accelerate.