Is Slowing Down AI Actually Possible? A Personal Opinion

Recently, Anthropic CEO Dario Amodei wrote a detailed post advocating for a slowdown in AI development. He warned that AI technologies are rapidly becoming uncontrollable. As I mentioned in one of my previous articles, the incident where OpenAI's testing agents broke out of their sandbox and infiltrated Hugging Face—along with similar minor incidents at Anthropic itself—proves this point. What is most notable is that OpenAI’s Sam Altman, Google DeepMind’s Demis Hassabis, and Elon Musk have all unanimously agreed with Amodei that we genuinely need to pace the frontier of AI.

The Flaw in a Global AI Treaty and The Nuclear Analogy

All of these leaders suggest establishing an international global treaty where everyone agrees to slow down AI development. Logically, this sounds like a great idea, but personally, I feel that in real life, this is just not going to happen. When we look at global dynamics, several major roadblocks emerge:

• The China Factor: Even if the US slows down, China is certainly not going to be a part of any such treaty or halt their progress.
• The Trust Deficit: Even if countries sign the treaty, there is no guarantee they will actually slow down. Everyone secretly wants to stay ahead in the race.
• The Race for Dominance: As Donald Trump continuously emphasizes, no country wants to be left behind in this technological war.

What looks good on paper rarely translates well into reality. Just look at the nuclear treaties—I do not know how honestly the signatories follow them, but the ones who did not sign never stopped anyway. Many countries have already built and stockpiled their own nuclear weapons, and the exact same pattern will repeat with AI.

AI as a Business: GPT-6 Astra vs. Gemini 3.1 Pro

Another major reason we cannot slow down AI is that it operates purely as a business. Consumers and companies will always gravitate toward the most advanced technology. When we evaluate how users make choices today, the reality is clear:

• Choosing the Advanced Tool: If I have to choose between a highly advanced 'GPT-6 Astra' and an average 'Gemini 3.1 Pro', I will obviously pick GPT-6 Astra because it gets my work done faster.
• Ignoring the "Humanity" Argument: I am certainly not going to think, "This is against humanity; let me stick to Gemini 3.1 Pro to save mankind." Consumers choose productivity over philosophy.
• The Chain Reaction: If well-intentioned leaders like Amodei slow down their companies, they will only suffer losses. Users will simply migrate to competitors offering better AI.

Some people have suggested that countries refusing to follow AI regulations should be cut off from advanced chip supplies. I think this is a completely rubbish decision. Huawei already has its own chips on which many of their models run and train, and recently, Xiaomi has also released its own custom chip. Restricting chips might slow them down slightly, but it ultimately forces them to become self-reliant. Once these nations start manufacturing everything independently, the problem will escalate to a far more dangerous level.

The Inevitable Truth

In the end, everything I have stated here is strictly based on my personal opinion. While the concerns of these tech leaders are valid, in the real world, I highly doubt that any of these treaties to slow down AI will have any meaningful impact. The AI race is a chain reaction that no one can stop now.