HuaRenCa
Back to Forum
Community

261106 NYT Profile of Dario Amodei: From a Physics/Biology Incel on Immersion Forums to the CEO of What Could Be the Largest IPO in Human History

梦萌
梦萌

2 weeks ago

Disclaimer: This article is a puff piece designed to build hype for the ANT IPO

Paywall-free version: https://www.nytimes.com/2026/11/05/technology/dario-amodei-anthropic-ai.html?smid=url-share&unlocked_article_code=1.OFE.CslN.usTW7cxPf_1Y

2007, Princeton University.

Dario Amodei, then 24, was pursuing a PhD in physics. He spoke with Colin Heckman, a high school classmate who was visiting, about Ray Kurzweil's 2005 book The Singularity Is Near.

Kurzweil predicted that artificial intelligence would surpass human intelligence at some critical point—the so-called "technological singularity."

Amodei believed that moment could arrive within their lifetimes.

Heckman recalled: "He was fascinated by the theory. He thought that if AI didn't develop in the right direction from the start, it could have catastrophic consequences for humanity."

At the time, Amodei read widely on artificial intelligence, technological development, and long-term risk, including blogs such as Accelerating Future and Overcoming Bias. He was also active in the online community of the charity evaluator GiveWell, taking part in discussions about how to maximize charitable impact.

The New York Times interviewed more than twenty of Amodei's classmates, friends, teachers, and former colleagues in an effort to reconstruct how these early experiences shaped his thinking, and how those ideas influenced his later founding and leadership of Anthropic.

Today, at 43, Amodei is CEO of Anthropic, one of the world's leading AI companies, which is also seen as a potentially major IPO candidate.

I. Rationalism and Effective Altruism

According to those interviewed, Amodei's views on technology were shaped mainly by two schools of thought: Rationalism and Effective Altruism.

The former emphasizes understanding the world through logic and rational analysis. In the AI field, some rationalists, represented by Eliezer Yudkowsky, argue that artificial intelligence surpassing human capabilities could pose an existential risk.

The latter advocates using quantitative analysis to compare the costs and benefits of different actions in order to maximize overall social welfare as much as possible.

Together, these two schools of thought shaped Amodei's judgment about AI development: artificial intelligence holds enormous potential, but it can also produce serious risks, so developers must establish corresponding safety mechanisms as early as possible.

According to friends and colleagues, he believes that since AI development is difficult to stop, responsible researchers should actively participate in technological development rather than leave it entirely to competitors who do not prioritize safety.

In 2021, Amodei co-founded Anthropic with his sister Daniela Amodei and five other researchers. Several of the founders had previously been connected to varying degrees with the rationalist or effective altruist communities.

As Anthropic has grown, Amodei has also become increasingly involved in public policy debates.

On international competition, he supports U.S. restrictions on exports of advanced AI chips to China in order to maintain America's technological advantage.

On AI governance, he advocates stronger regulation and opposes the use of Anthropic's technology for large-scale domestic surveillance in the United States or for fully autonomous combat weapons.

These positions have drawn mixed reactions in the industry.

Supporters argue that his emphasis on AI risk helps promote responsible technological development. Critics argue that overemphasizing risk can become a justification for restricting competition and concentrating control over technology.

Meta founder Mark Zuckerberg and Nvidia CEO Jensen Huang have both criticized similar narratives about AI risk.

The U.S. Department of Defense has also publicly responded to Anthropic's stance, saying that what it pursues is "Americanism, not effective altruism."

In recent years, Amodei himself has tried to keep his distance from particular schools of thought.

In 2023, he said on The New York Times podcast Hard Fork that he agreed with some of the ideas of effective altruism but did not want to be categorized into any ideological camp.

"I do agree with a lot of the ideas in it, but I oppose the tribal thinking that insists on sorting people into different camps."

II. Upbringing

Dario Amodei and his sister Daniela grew up in San Francisco's Mission District.

His father, Riccardo Amodei, was a leatherworker, and his mother, Elena Engel, worked on public library renovation projects. According to divorce records, his parents separated in 1991.

Amodei showed strong academic interests from an early age.

In 1997, when he graduated from Lawton Alternative School, he and another classmate were voted "most likely to succeed."

He later attended Lowell High School in San Francisco and took courses at UC Berkeley ahead of schedule.

High school classmates recalled that he was already studying utilitarianism at the time—the ethical theory that seeks to maximize overall welfare by comparing the utility of different outcomes.

He also followed the Y2K issue, worried that computer systems might suffer large-scale failures as a result.

His high school physics teacher, Richard Shapiro, recalled that Amodei held himself to very high standards on exams.

"I often told him: Dario, don't be so nervous!"

In 2001, Amodei entered Caltech.

During college, he often returned to San Francisco to meet high school friends and discuss ethics, social issues, and technological development.

At one dinner with Colin Heckman and Nathan Avery, the group discussed a utilitarian thought experiment:

Suppose someone is slaughtering chickens in huge numbers. Under what circumstances would killing that person to save more chickens be justified within a utilitarian framework?

According to Avery, Amodei believed that if the number was large enough, the total value of the chickens' lives could in theory exceed the value of one human life.

That answer reflected his interest at the time in utilitarian reasoning.

Amodei describes himself as an atheist and was also actively involved in social issues. In 2003, he joined friends in a protest against the U.S. invasion of Iraq.

In 2004, he transferred to Stanford University, partly because he wanted to be closer to his sister Daniela, who was attending high school nearby.

The siblings have always been close.

In 2023, Amodei said on a podcast: "We always wanted to save the world together."

III. From Physics to Artificial Intelligence

In 2006, Amodei graduated from Stanford and entered Princeton University to pursue a PhD in physics, gradually shifting his research toward biophysics and computational neuroscience.

That same year, his father died of hepatitis C.

Afterward, he began studying retinal cells while also thinking about how to develop general tools that could accelerate scientific discovery.

His former girlfriend Jade Wang recalled that the two once co-organized a science discussion group and often discussed one question:

Compared with studying a single disease, could developing methods and tools that can be widely applied across multiple research fields produce greater social value?

This line of thinking was connected to Amodei's family experience. When his father died, effective drugs capable of curing most hepatitis C patients had not yet become available.

At the same time, he began reading deeply the works of Kurzweil and the AI rationalist thinker Eliezer Yudkowsky.

Among them, Kurzweil's transhumanist ideas had a relatively large influence on him. That theory holds that technological progress may ultimately help humanity break through the limits of physiological and cognitive ability.

In 2007, Amodei predicted to Heckman that the technological singularity might arrive before 2040.

He envisioned that humanity might in the future develop into superintelligent AI, or form new forms of intelligence through human-machine integration.

He also actively recommended that friends and his sister attend the annual conference of the Singularity Institute founded by Yudkowsky. The organization was later renamed the Machine Intelligence Research Institute (MIRI).

A 2008 email reviewed by The New York Times shows that Amodei once called Yudkowsky "one of the deepest thinkers I have ever met" and praised his efforts to promote a technological singularity beneficial to humanity.

However, in response to The New York Times, Yudkowsky said he did not remember Amodei ever seeking his advice or adopting his suggestions.

Around this period, Amodei also began corresponding with Elon Musk.

According to two people familiar with the matter, he expressed admiration for Musk and considered joining SpaceX, but ultimately did not.

IV. Effective Altruism and Life Sciences Research

In 2008, Amodei learned about the charity evaluator GiveWell through economist Tyler Cowen's blog.

After that, he began participating in GiveWell's email discussions and forum activities, studying how to allocate charitable resources more effectively.

In 2010, he said in an email that he was considering entering the finance industry in order to earn more income for charitable donations.

That same year, he wrote for GiveWell about how to allocate a $10,000 donation.

He compared two projects:

Stop TB Partnership, which focuses on preventing and treating tuberculosis; and VillageReach, which mainly supports childhood vaccination in rural Africa.

In his initial analysis, Amodei believed that saving adults might produce greater utility, because adults usually have richer experiences and social relationships.

He wrote: "I think the tragedy of an adult death is probably two to three times that of an infant death."

But after further comparing the projects' operational efficiency, he concluded that VillageReach could use funds more effectively and ultimately chose the latter.

This experience reflected his emphasis on results-oriented and quantitative evaluation methods.

In 2011, Amodei went to Stanford University for postdoctoral research, focusing mainly on cancer cells.

Friends recalled that he was very focused on health, aging, and time management.

He kept up regular swimming, tried anti-aging supplements, and studied the potential effects of calorie restriction. At the time, animal experiments had already found that restricting calorie intake might extend the lifespan of certain laboratory animals.

High school friend Avery recalled that Amodei often discussed the finiteness of life and believed that time is a resource that cannot be regenerated.

In 2014, Amodei was diagnosed with testicular cancer and later recovered after chemotherapy.

According to two people familiar with the matter, this experience further strengthened his sense of time pressure.

At the same time, he developed close ties with GiveWell co-founder Holden Karnofsky.

At the time, GiveWell was working with Good Ventures, founded by Facebook co-founder Dustin Moskovitz and his wife, to support effective altruism-related projects.

Amodei served as an adviser to GiveWell Labs, mainly advising on funding allocation in the life sciences.

Karnofsky was then dating Daniela Amodei. The two rented a house in San Francisco, which gradually formed into a shared-living community of researchers and friends.

Later members who moved in included Dario Amodei, theoretical physicist Jared Kaplan, and Christopher Olah, who had once received math tutoring from Amodei.

According to people who visited the residence, the roommates often discussed Yudkowsky's writings, as well as rationalism, effective altruism, and the future of artificial intelligence.

They also attended Burning Man in Nevada.

This period of living together also led to long-term research and collaborative relationships.

Kaplan and Olah later co-founded Anthropic with the Amodei siblings.

In 2025, Holden Karnofsky also formally joined Anthropic.

V. Baidu, Google Brain, and OpenAI

In the early 2010s, deep learning began making major advances, and technology companies ramped up investment in AI research.

Baidu was one of the technology companies that began large-scale deep learning research relatively early.

Baidu established a research institute in Sunnyvale, California, and hired former Google researcher Andrew Ng as chief scientist.

In 2014, Dario Amodei joined Baidu's AI research team.

Colleagues recalled that he was dedicated to his work, often staying in the office late into the night, while also holding distinct views on technology and research directions.

Gregory Diamos, a researcher who worked with him, said: "Dario seized almost every opportunity for intense debate with people."

Diamos recalled that even then, Amodei already showed considerable concern about the long-term safety of AI.

In 2015, he told colleagues that AI capabilities might reach or even surpass human levels in a relatively short period of time, so research institutions should consider the risk of technology going out of control in advance.

That same year, Amodei left Baidu and joined Google Brain, beginning to focus more on AI safety research.

In 2016, he and five other researchers, including Christopher Olah, co-published the paper "Concrete Problems in AI Safety."

The paper discussed a range of practical safety problems that AI systems might encounter, including poorly designed reward mechanisms, unintended behavior, and the ability of systems to adapt to environmental changes.

That same year, Amodei joined OpenAI, which was still a nonprofit research organization at the time.

At OpenAI, he was responsible for safety research while also leading a team studying Scaling Laws—how model performance changes as training data, computing resources, and model size increase.

This line of research later became an important foundation for the development of large language models.

According to seven former colleagues, during his time at OpenAI, Amodei gradually developed a clearer technology development roadmap and actively pushed to incorporate safety research into core R&D work.

He also began linking AI development to the political environment.

In November 2016, three days after Trump first won the U.S. presidential election, Amodei expressed concern in an internal OpenAI communication about the direction of U.S. politics.

According to internal records reviewed by The New York Times, he wrote that he could not accept participating in the development of frontier technology with major military potential in an authoritarian country.

He also criticized investor Peter Thiel for donating $1.25 million to Trump's campaign, arguing that OpenAI should stop accepting Thiel's money.

However, Thiel's Founders Fund later became one of Anthropic's investors.

At the same time, Amodei also helped OpenAI seek funding aligned with its research goals.

In 2017, he introduced OpenAI to the charitable foundation led by Holden Karnofsky.

The organization, now called Coefficient Giving, subsequently donated $30 million to OpenAI.

Karnofsky later joined OpenAI's board.

That same year, Karnofsky married Daniela Amodei.

According to the wedding website and photos from the event, the wedding had a relaxed theme, with guests taking part in a stuffed-animal matching game and a dinner requiring "funny formal wear."

Dario Amodei attended wearing a panda head mask, while Christopher Olah wore a giraffe head mask.

Amodei's then-girlfriend Cami Clark was also there. The two married in 2022.

VI. Divisions Inside OpenAI and the Founding of Anthropic

In 2018, Daniela Amodei joined OpenAI and later rose to vice president in charge of safety and policy.

As the siblings' influence inside OpenAI grew, some employees began to worry about the concentration of decision-making power over research.

According to three people familiar with the matter, in December 2018, OpenAI researcher Joshua Achiam expressed doubts to CEO Sam Altman about appointing Dario Amodei as chief ethics officer.

Achiam believed that AI ethics and governance decisions should involve people with different backgrounds and viewpoints, rather than being overly concentrated in the hands of a few. He suggested that an outside expert hold the position.

According to people familiar with the matter, Amodei strongly opposed the suggestion and believed he was the most suitable candidate. The two sides then had a heated argument.

Also in 2018, Dario and Daniela Amodei began leading an internal OpenAI research project code-named GPT.

According to six former employees, while advancing research on model capabilities, the team also explored how to use safety alignment techniques to make AI better follow human instructions.

This direction later became one of Anthropic's core research principles: while continuously improving model capabilities, develop safety technologies to match.

However, control over the GPT project also became a focus of internal conflict at OpenAI.

According to two people familiar with the matter, then-OpenAI president Greg Brockman wanted to be more involved in the project but was opposed by the Amodei siblings.

Dario Amodei repeatedly told Sam Altman that if Brockman continued to intervene in the related work, he would consider resigning.

According to a person familiar with Altman's thinking, Altman ultimately did not accept the demand.

Divisions over research direction, management authority, and organizational governance gradually deepened.

In December 2020, Dario Amodei, Daniela Amodei, and several core researchers left OpenAI.

About four weeks later, they co-founded Anthropic.

The company's founding both continued Amodei's long-standing focus on AI capabilities and safety issues and reflected his disagreements with OpenAI management over technology strategy and decision-making mechanisms.

From his early interest in the technological singularity to effective altruism, life sciences research, and then his experiences at Baidu, Google Brain, and OpenAI, Amodei gradually formed a relatively stable set of views on technology:

Artificial intelligence may have profound social impact, and its capability development is difficult to stop, so safety research should advance in step with technological progress.

This idea became an important foundation for Anthropic, and it also keeps Amodei at the center of ongoing controversy over AI regulation, technological competition, and governance power.

4
1.2K

Comments (4)

Your avatar
Sign in to comment
LKR-洛凯瑞
LKR-洛凯瑞2 weeks ago

Reconstructing motives from interviews and old emails alone tends to reduce complex people to a single line of thought.

CC123
CC1232 weeks ago

The core judgment is actually quite clear: his safety rhetoric and his commercial expansion are two sides of the same logic. Talking about risk in public both helps him win a voice in policy debates and raises the bar for those who come after. When reading profiles of figures like this, it's worth comparing the funding timeline against his regulatory statements—that's more useful than just looking at his views.

cacty
cacty2 weeks ago

Predicting in 2007 that the singularity would arrive before 2040 — that timeline itself is worth pondering. The window left isn't wide, which also explains the sense of urgency.

r961961r
r961961r2 weeks ago

Look at it from another angle: instead of arguing over which school of thought he belongs to, it's more useful to see where he actually directs his resources.