Summary of Key Points
Dario Amodei, the founder of Anthropic, is a man filled with contradictions. On one hand, he loudly warns about the potential disasters brought by super AI, such as biological weapons and mass unemployment, while on the other hand, he leads his company in the frantic development of cutting-edge AI models. He believes that AI is inevitable, and rather than letting others control it, he wants to take the reins himself. He attempts to ensure safety through concepts like “constitutional AI” and trust structures, but he also finds himself in a dilemma of holding both the position of athlete and referee, concentrating power in his own hands. He dreams of AI curing cancer and benefiting all of humanity, yet beneath this lies his ambition for American hegemony. There are even rumors that he aims to make Anthropic the “last private company in the world.” Although he avoids direct confirmation, his core logic remains clear: only by holding power can things be trusted.
Dario's Contradictions: Warning of Danger While Accelerating Progress
Dario’s contradictions are not feigned; they stem from two obsessions:
1. The Gene of Speed: His father died of hepatitis C when a new drug was still 7 years away from approval. This experience made him believe that “slowness is a sin” and that science must move fast to save more lives.
2. The Prisoner Dilemma of the Industry: He understands the dangers of AI, but companies like OpenAI and Google are also racing to develop it, and other countries will not stop. If Anthropic pauses, someone less trustworthy may create super AI first. Therefore, his logic is: “Since everyone is working on it, I might as well be the one to do it first so that I can control it.”
He lists 100 dangers of AI in his articles while simultaneously using AI to write code and accelerate development for Anthropic. To him, this is not a contradiction; it’s a “necessary risk” – only by gaining control can safety be ensured.
Anthropic’s Safety Measures: Taking Control for Peace of Mind
Dario left OpenAI because the people who developed the models there didn’t have final say. After founding Anthropic, he embedded the concept of control into the company’s DNA:
- Constitutional AI: He created a set of “values” for the models to follow, which they use to check their own responses against these principles. However, this “constitution” was defined by Anthropic itself – essentially, a single company deciding what characteristics AI should have on behalf of all of humanity.
- Long-Term Interest Trust: He formed an external trust to theoretically prevent the company from sacrificing safety for profit. Yet, the power of the trust still revolves around Anthropic, meaning it’s “self-supervising.”
- Being Both Athlete and Referee: Anthropic develops models, assesses risks, sets standards, and provides recommendations to regulators—all done in-house. It’s like a player also being the referee; do you really believe they will be fair?
His view of safety boils down to: “Only I understand AI, so only I can manage it.”
The Future of AI with American Priority: Ambition for Hegemony
Dario claims that AI could cure cancer and help poor countries skip the industrialization process, but his vision of benefiting humanity is predicated on the United States gaining control of AI first:
- He advocates that the US and its allies should block advanced chips to prevent strategic competitors like China from accessing the resources needed to train the strongest models.
- He uses the metaphor of “an eternal 1991” to describe his ideal future: a world where the West, after the dissolution of the Soviet Union, dominates with AI and then shares its benefits with others.
- When Chinese company DeepSeek creates competitive models with limited resources, he argues that this proves China’s strength and further justifies chip restrictions – a typical logic of “the stronger the opponent, the more you need to block them.”
His so-called “common good” is actually a plan for the US to take control first, with Anthropic managing everything afterward.
The “Last Private Company” Myth: The Ultimate Power Dream
There are rumors that Dario wants Anthropic to become the “last private company in the world,” leaving only it and governments. He hasn’t responded directly, but his actions speak volumes:
- He has eliminated all “untrustworthy” parties: the market (driven by profit), the public (lacking technical understanding), other companies (not cautious enough), and non-Western governments (authoritarian). In the end, only Anthropic remains.
- When responding to these rumors, he mentions curing cancer as a more compelling goal, thereby avoiding the core issue of power concentration.
This reminds us of the Pharaoh in “The Watcher”: he sees himself as a savior, yet for the sake of “peace,” he brings about disaster. Dario also believes he is the only one who can save humanity, but he doesn’t realize that handing all power to one company represents the greatest danger.
Conclusion
Dario is a sincere idealist, but sincerity does not equate to correctness. He wants to protect the world by controlling AI, yet in doing so, he gradually becomes the center of that power. Just as the ending of “The Watcher” suggests, “Nothing is over; everything will never end.” When a company believes it can make decisions on behalf of all of humanity, it may have already become the very “abyss” it warns against.