This article is based on a feature from The New York Times on November 5, titled "The Making of Anthropic's C.E.O." (by Ryan Mac, Kalley Huang, Cade Metz, Georgia Gee)
This month, Anthropic is going public, set to be one of the largest IPOs in the world. The New York Times released a profile of Dario Amodei, interviewing over twenty of his classmates, friends, teachers, and colleagueswhile Dario himself declined the interview
As the founder of the company, Dario is 43 years old this year, and chronologically, he has experienced the following:
In 2006, his father died of hepatitis C
In 2007, after reading "The Singularity Is Near," he concluded that the singularity would arrive around 2040
In 2014, he was diagnosed with testicular cancer and fully recovered after chemotherapy
In 2020, he fell out with Altman, left OpenAI, and founded Anthropic a few months later
Friends and colleagues mentioned that Dario's ideas primarily stem from two schools of thought: rationalism and effective altruism (EA), both of which believe that mathematical reasoning can change the future of humanity
Thus, Dario formed the belief that:
AI is too dangerous to be created by anyone other than him and like-minded individuals, so it needs to be developed before those who do not care about safety do, and kept in his own hands
He Values Time Most
Dario's obsession with time began with his father. In 2006, his father passed away from hepatitis C, and just a few years later, new medications emerged that could cure most hepatitis C patients
That year, Dario had just graduated from Stanford and went to Princeton to pursue a PhD in physics. After his father's death, he gradually turned towards biophysics and computational neuroscience, pondering how to accelerate science with technology
His then-girlfriend Jade Wang recalls that they often discussed a question: who has a greater impact, a scientist who spends a lifetime studying a disease, or someone who creates tools for the entire field? Dario chose the latter, even using his father as an example
By 2011, he returned to Stanford for a postdoc, researching cancer cells
At the same time, he began taking anti-aging supplements, maintained a strict swimming regimen, and also studied calorie restriction (some studies have shown that this can extend the lifespan of laboratory mice)
A high school cross-country teammate, Nathan Avery, said Dario often remarked that time is the only non-renewable resource
Then in 2014, he himself was diagnosed with testicular cancer, underwent chemotherapy, and later fully recovered
Two people who knew him said this event made him even more urgent about time
In the same year, he joined Baidu's newly established AI lab in Silicon Valley
Putting Morality into Numbers
Dario grew up in the Mission district of San Francisco and attended the best public high school in the area, Lowell. In his spare time, he would take classes at UC Berkeley
At that time, he was already contemplating utilitarianism: quantifying the consequences of each choice and selecting the one with the highest sum
While in college (first at Caltech, then transferring to Stanford), there was one time he went for sushi with friends. He was vegetarian, but argued that crabs are invertebrates and therefore acceptable to eat
A friend then posed a question: if someone continued to kill chickens, at what point would it be justified to stop them by killing that person?
Dario chuckled and acknowledged that there is indeed such a number
Later, he applied this method of calculations to his donations
In 2008, he came across GiveWell on economist Tyler Cowen’s blog, an organization researching how to maximize the impact of every donation, and he quickly became an active user of the forum
In 2010, he wrote an article for GiveWell discussing who his $10,000 donation should go to. There were two candidates: one for preventing and treating adult tuberculosis, and the other for vaccinating children in African villages
Initially, he calculated that an adult could have deeper and more meaningful experiences, so the death of an adult would likely be two to three times worse than that of an infant, suggesting he should choose the former
But in the end, he compared the execution situations of both organizations and chose to donate to the one providing vaccinations
In the same year, he also seriously considered going into finance, as earning more would allow him to donate more
Predicting the Singularity in 2040
As for AI, Dario became fascinated in 2007. At that time, he was 24 years old, studying for his PhD at Princeton, and had just finished reading Ray Kurzweil's "The Singularity Is Near." A high school classmate, Colin Heckman, came to visit him, and he told Heckman that the singularity would arrive around 2040, at which point humans might merge with machines, becoming new superintelligences
Heckman said Dario became obsessed from that point on, firmly believing that AI must be created correctly; if created incorrectly, humanity was doomed
During those years, he frequently read the blogs "Accelerating Future" and "Overcoming Bias," where the authors often discussed how AI might destroy humanity
He also greatly admired the representative figure of rationalism, Eliezer Yudkowsky. In 2008, he mentioned in an email to a friend that Yudkowsky was one of the most profound thinkers he had encountered, and he brought friends and his sister to attend the annual conference held by Yudkowsky's Singularity Institute (now called MIRI)
This time, Yudkowsky responded to reporters by saying he did not remember Dario asking him for any advice
(By the way, during this period, Dario also communicated with Musk through letters, considering working at SpaceX)
The Team Came from a Shared House
As a background, during those years, Dario was very close to Holden Karnofsky, the founder of GiveWell, and served as a life sciences advisor to GiveWell Labs (a project co-hosted by GiveWell and the charitable organization founded by Facebook co-founder Dustin Moskovitz and his wife)
Karnofsky was dating Daniela at the time, and they moved into a shared house in San Francisco. Later, Dario, theoretical physicist Jared Kaplan, and Chris Olah, who had studied math with Dario, also moved in
The roommates would read Yudkowsky's articles together, attend Burning Man together, and sometimes dress up in animal costumes
Four of the seven co-founders of Anthropic lived in this house
The other three also spent time in the rationalism and EA circles
As for Karnofsky, he married Daniela in 2017 (at the wedding, Dario wore a panda head), and joined Anthropic in 2025
Fighting for Control at OpenAI
From Baidu onwards, Dario became known for his love of debate
The Baidu lab was led by Andrew Ng, who had come over from Google. Colleague Gregory Diamos recalls Dario sparking discussions at almost every meeting. As early as 2015, he warned colleagues that AI could quickly catch up to or even surpass the human brain, and that companies involved in AI needed to take the possibility of losing control seriously
In the same year, he joined Google Brain, specifically focusing on safety, and in 2016 published the paper "Concrete Problems in AI Safety" with Olah and others
Later, he joined OpenAI, leading the safety team and the scaling laws team. Seven former colleagues said that during his time there, he became increasingly bold in voicing his opinions and more determined to push the company according to his own ideas, such as:
Three days after Trump's election in 2016, he expressed in an internal chat that he was worried about America slipping into authoritarianism and was unwilling to research militarily potent technologies in such a country
He also stated that Peter Thiel's funding of Trump's campaign was morally reprehensible and that OpenAI should no longer accept his money (later, Thiel’s Founders Fund invested in Anthropic)
In 2017, he facilitated a $30 million donation from the charity that Karnofsky chaired (now called Coefficient Giving) to OpenAI, after which Karnofsky joined the board
In 2018, researcher Joshua Achiam advised Altman not to make Dario the Chief Ethics Officer, suggesting that external experts would be a better fit. Dario raised his voice at him, saying no one was more suitable than himself
In the same year, he and Daniela took over the GPT project, attempting to teach AI to accomplish tasks under human guidance. Initially allowing the model to approach dangerous capabilities, then using safety technologies for verification, which later became his consistent approach
Greg Brockman wanted to be involved in GPT; Dario repeatedly gave Altman an ultimatum: if Brockman wasn't kept out, he would leave
Altman refused
In December 2020, the siblings and several colleagues left OpenAI and founded Anthropic shortly thereafter
(By the way, The New York Times is suing OpenAI and Microsoft for copyright infringement, both of which deny the allegations)
Every Stance Has a Reason
Dario's current public positions are all related to the risks of AI:
Limiting U.S. chip sales to China
Regulating AI to prevent it from spiraling out of control
In disputes with the U.S. Department of Defense, insisting on drawing two lines for AI’s applications: large-scale surveillance of Americans and autonomous weapons
In September this year, he published an article calling for slowing down the development of frontier AI
Many in the AI safety community support him, while Zuckerberg and Jensen Huang have criticized him as a "doomsayer," claiming he wants to hold power over AI
The U.S. Department of Defense also posted on social media in September:
"Americanism, not effective altruism."
Dario himself is also keeping a distance from both circles. In 2023, during the Hard Fork podcast, he stated that he does not consider himself EA, recognizing many of its ideas but opposed to dividing people into tribes
As for why he insists on taking this position, high school cross-country teammate Avery said:
Dario genuinely believes that humanity faces immense survival risks, and therefore he must place himself in a position to do something about it
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到[email protected]，本平台相关工作人员将会进行核查。
