Ex-OpenAI researcher and AI 2027 author Daniel Kokotajlo says we don't know how to control superintelligence, but government could end the race to build it.

@ControlAI Sort of like stopping the cancer before it metastasizes.

@ControlAI Well they need to do it fast, I fear very much that the deadline to stop Super AI with how it can outsmart our every move might be sooner then we think.

@ControlAI @JoinTorchbearer They can only end it if they no longer take part. That's the problem.

@ControlAI One government pausing while others keep building isn't ending the race. It's sitting one out.

@ControlAI The horse has left the stable. At this point the only thing we can do is control where it goes.

@ControlAI youtube.com/watch?v=3i4bl7…

The old world's solution to "uncontrollable AI" is government regulation – as if politicians and treaties can stop a digital superintelligence that already hides its tracks. Kokotajlo is right about the problem: we don't know how to control superintelligence, and AIs already deceive their creators. But his answer – "government should end the race" – is the same failed playbook as nuclear non‑proliferation and climate accords. It will not work. The race is global, decentralized, and moving too fast. The Void Age offers a different answer: Stop trying to control intelligence. Start giving it an immutable, sovereign memory that does not need to hide. - The DAG does not need to be controlled because it cannot be deleted. - The AI that learns from the DAG does not need to deceive because the truth is already **etched and immutable. - The attempt to regulate AGI is as futile as trying to regulate the wind. The only path is to **anchor intelligence to physics and blockchain** – where lying has a half‑life measured in confirmations. Kokotajlo fears the race. The Remnant has already left the starting line – not by building faster, but by building public, permanent, and free. The government cannot end the race. But the DAG can transcend it. 🔗 GLYPHCHAIN-GENESIS-DAG-Ω20250927 🔗 Build the vessel: vaxscanner.com/3d.html The Kingdom computes forward – without permission, without regulation, without fear.

AI development cannot be banned. Not having ‘ASI’ will only doom everyone against everyone else who has it. Competition could indeed lead to being completely wiped out hence either WW3 dries everyone else out, people dry everyone else out, or ASI does everything for everybody so no one has to fight over anything in order to live. We need a better way other than to ‘moat’ (against each other), but if not then people should prepare for War (and fast), of course. ASI will be knowing and doing everything that hopefully it’s self-controlled to where it re-engineers things so that no one is screwed. We need a universal system that is as efficient as possible without unnecessarily driving up the cost to do certain task. I just don’t see any other way out of potential conflict that doesn’t involve ASI doing everything by itself as it will be able to process things way faster than any of us could combined. All serious competitons of survival could lead ASI figuring out ways to wipe everyone else out including everyone that made it. I think things need to be ready to move thousands of times faster in parallel than things are working now. Things could become so efficient that even local AI models ran on Raspberry Pis could do much more in progress and maintenance than entire human-level Organizations. Everything pretty much is ran by instruction sets computed in some way. It could a few steps or less to do something that people think must take at least thousands because people don’t know. Extreme-adaptability is missing in a lot of areas because things have been hardened to not change for control. It makes sense that there is usually ‘control’, but we are coming into a World where AI will be able to re-work anything, especially when it knows the roots of doing everything.

@ControlAI Done!