A deep dive into the founding of DeepMind, OpenAI, and Anthropic, as many of their backers feared the risks of advanced AI but wanted the power to shape it
The people who were most afraid of the risks of artificial intelligence decided they should be the ones to build it.
New York Times
Context & Ripple Effects
The founding narratives of DeepMind, OpenAI and Anthropic share a central tension: funders concerned about advanced-AI risk sought influence by helping build the organizations advancing the technology. That tension was later made explicit in a cross-industry call to treat extinction risk as a global priority.
The safety rationale has not removed disputes over how frontier labs are run. Reporting on staff warnings about secrecy and recklessness shows that internal governance and disclosure became part of the argument over whether safety-oriented institutions can credibly deliver on their mission.
First-order effects
The account gives DeepMind, OpenAI and Anthropic a shared founding logic: safety concerns were not only a case for restraint, but also a reason for backers to seek a seat inside frontier-AI development.
It puts the governance choices of these labs under sharper scrutiny, because their safety claims are tied to who financed and shaped them from the outset.
Second-order effects
Rival labs and their investors face pressure to explain whether safety commitments constrain deployment decisions or principally justify control over powerful models.
Employees, critics and policymakers gain a clearer basis for judging conflicts between a lab's stated risk mission and its commercial or strategic incentives.
Third-order effects
If risk-conscious capital continues to fund the most capable labs, AI safety may increasingly be governed through concentrated private institutions rather than external rules alone.
The durable question becomes whether private stewardship can remain accountable as frontier labs compete for capital, talent and influence; the later debate over Anthropic's independence underscores that unresolved trade-off.
The trend: Frontier AI is being shaped by a paradox in which fears of concentrated technological power motivate investors to concentrate influence inside the labs building it.
By the way all the hagiography about OpenAi & Musk is annoying. He apparently thought they were goin nowhere. So he bailed after poaching a top engineer/researcher over to Tesla. That guy, Andrej Karpathy, later quit Tesla & returned to OpenAi. Now, OpenAi is huge & Tesla may…
The existential risk comes from the goal of wanting to replace human decision making with computers. If your goal is to have LLMs running the stock market, health insurance system, criminal justice or military weapons then you have an existential risk. …
I think there's something attractive to techies of framing the OpenAI drama as noble hearted researchers wanted to do charitable work “for humanity” and greedy business people wanted to monetize their ideas. In a world where working in big tech is synonymous with selling out, it…
This is an extraordinary work of narrative journalism about what's been happening the last several years while most of us were focused on other things. https://www.nytimes.com/...
Jeesh. An artificial intelligence writes about artificial intelligence as Dowd sides with Musk over AI... Sam Altman, Sugarcoating the Apocalypse https://www.nytimes.com/...
OpenAI's Greg Brockman to @maureendowd , 2016; “It's not enough just to produce this technology and toss it over the fence and say, ‘OK, our job is done. Just let the world figure it out.’” OpenAI to the world, with it's GPT-4 system card, 2023: Yeah, this could be really bad,...
This is just one incredible detail from the massive heave on the history of modern AI from my colleagues on the NYT tech desk. In 2017, Elon Musk sought to push Open AI to go faster and pursue the exact commercialization that he's criticizing today. https://www.nytimes.com/... [i…
Fantastic anecdote «Mr. Musk explained that his plan was to colonize Mars to escape overpopulation and other dangers on Earth. Dr. Hassabis replied that the plan would work — so long as superintelligent machines didn't follow and destroy humanity on Mars, too. Mr. Musk was spee…
The “masters of the universe” approach to tech coverage is never the real story. It is the made for Hollywood version but it miss so much. Sad to see it coming back in style. https://www.nytimes.com/...
Good NYT survey of history of AI in silicon valley... mostly right by my understanding with maybe a few bits i have heard differently. don't miss key callouts around “people who are most worried also trying to build it and profit” and “cult analogies” [image]
to understand how we got to where artificial intelligence is top of the tech agenda for every megacorp and geopolitical body today, you need to go back... Ego, Fear and Money: How the A.I. Fuse Was Lit by @CadeMetz, @KYWeise, @nicoagrant and me https://www.nytimes.com/...
What this NYT article tells us about the genesis of the new AI industry is that there was a lot of drama at Google, DeepMind, OpenAI, and Anthropic. Comparatively, the creation of FAIR was straightforward and drama free. https://www.nytimes.com/...
“Mr. Page insisted his utopia should be pursued. Finally he called Mr. Musk a ‘speceist’. That insult, Mr. Musk said later, was ‘the last straw.’” MUST READ: Epic piece by @CadeMetz @KYWeise @nicoagrant @MikeIsaac https://www.nytimes.com/...