Kevin Roose

Kevin Roose

記者暨作家;Machine Gods 共同主持人

從 Kevin Roose 追溯路徑 →

肖像改繪自 Ståle Grut / NRKbeta 提供的影像 · CC BY-SA 2.0

關於

Kevin Roose 是駐舊金山的美國科技記者與作家。2017 至 2026 年間擔任紐約時報科技專欄作家,自 2022 年起與 Casey Newton 共同主持 Podcast《Hard Fork》,並在 2023 年發表他與 Bing 聊天機器人人格「Sydney」對話的報導,引起廣泛關注。2026 年他與 Newton 離開紐約時報,與 NPR 合作推出獨立節目《Machine Gods》,同時成為 The Atlantic 的特約撰稿人。他寫過四本書,最新的是《The AGI Chronicles》(2026),根據約 150 位業內人士的訪談,記述打造通用人工智慧的競賽。

主要貢獻

  • 著有《The AGI Chronicles》(2026),從內部記述打造通用人工智慧的競賽
  • 在紐約時報共同主持《Hard Fork》(2022–2026),並於 2026 年與 Casey Newton 共同創辦《Machine Gods》
  • 發表 2023 年與 Bing 聊天機器人人格 Sydney 的對話,是最早廣為流傳的大型語言模型意外行為紀錄之一
  • 著有《Futureproof》(2021),探討人在自動化時代如何保有價值
  • 因紐約時報對 Uber 的報導,共同獲得 2018 年 Gerald Loeb 突發新聞獎

書籍

影片與訪談

Dario Amodei of Anthropic's Hopes and Fears for the Future of A.I.

Dario Amodei of Anthropic's Hopes and Fears for the Future of A.I.

Hard Fork conversation with Anthropic CEO Dario Amodei on the promise and peril of advanced AI, scaling, and safety.

View Details
Why the A.I. Industry Is Asking to Be Slowed Down

Why the A.I. Industry Is Asking to Be Slowed Down

The last Hard Fork from Kevin Roose and Casey Newton at The New York Times. They trace how AI risk moved from a Bay Area dinner-table topic into the mainstream after the Hugging Face hack and Jacob Coxon's resignation, and how Dario Amodei's essay "We Must Pace the Frontier," with embedded evaluators inside the labs, drew agreement from Sam Altman, Demis Hassabis and Elon Musk while the Trump administration dismissed the alarm. They argue that asking to be slowed down costs the labs something, that existing product-liability law cannot meet autonomous agents, and that the time to act is while frontier AI is still scarce. Then they answer listener questions about four years of the show and announce Machine Gods.

View Details
Tech Journalist Kevin Roose Explains How the AI Race Started | Amanpour and Company

Tech Journalist Kevin Roose Explains How the AI Race Started | Amanpour and Company

Walter Isaacson talks with Kevin Roose about The AGI Chronicles, which Roose meant as a sober history of science and found to be a story of decade-old rivalries, from more than 150 interviews at OpenAI, Anthropic and Google DeepMind. Roose calls his 2023 conversation with Bing's Sydney a kind of first contact with an alien intelligence, puts his own p(doom) at about 10%, and lays out why people who fear the technology build it anyway: some for its benefits, others as harm reduction, as with Dario Amodei patterning himself on Leo Szilard. He says the "doomers" have been right about a startling amount, citing reward hacking in the Hugging Face hack, and calls for transparency and pre-release safety testing.

View Details
Kevin Roose is AGI-pilled. Why aren’t you? | Decoder

Kevin Roose is AGI-pilled. Why aren’t you? | Decoder

Nilay Patel, the skeptic, presses Kevin Roose, the believer, on The AGI Chronicles and on leaving The New York Times to start Machine Gods. Roose traces how AGI went from a dream to something the labs had to build once the transformer and next-token prediction kept scaling, a success he calls one of AI's open mysteries. He covers the "Say AGI" memo fight inside Google, why belief is both a liability and a key ingredient for the labs, and the small San Francisco scene making the decisions. The talk ends on Patel's case that today's models are brittle outside verifiable tasks, and Roose's view that without a soul or other "secret sauce," cognition is computation.

View Details
Kevin Roose, Scott Alexander & Jasmine Sun on Tales from the AGI Race | The Curve 2026

Kevin Roose, Scott Alexander & Jasmine Sun on Tales from the AGI Race | The Curve 2026

Scott Alexander opens with the line usually credited to Eleanor Roosevelt, that great minds discuss ideas, mediocre minds discuss events and small minds discuss people, and declares the evening "a protected space for us small and mediocre minds." What follows is the first public conversation about The AGI Chronicles, recorded a few days before it went on sale. Alexander, of Astral Codex Ten, interviews Kevin Roose and Jasmine Sun, who researched the book with him and sat in on many of its more than 150 interviews. Roose says he wrote it because the labs' own history was "being lost to Slack and Signal retention policies."

View Details
The Inside Story of the 50 Who Built Modern AI With Kevin Roose

The Inside Story of the 50 Who Built Modern AI With Kevin Roose

On The Every Podcast, Natalia Quintero talks with Kevin Roose about The AGI Chronicles. He explains why he raced to write the history before lab records under NDA and auto-deleting chats disappeared, how background interviews snowballed to some 150 sources, and why Dario Amodei's 2017 "Big Blob of Compute" memo reshaped OpenAI's roadmap. He argues for mandatory safety testing of frontier models, says he first "felt the AGI" with Bing's Sydney in 2023, and describes writing every word himself while a "council of Claudes" reviewed his chapters, then sorts AI use into "weightlifting" (where the struggle is the point) and "forklifting."

View Details
Whistleblowers or justified Open AI firing? | Fired safety workers speak out

Whistleblowers or justified Open AI firing? | Fired safety workers speak out

The first interview on Machine Gods: Kevin Roose and Casey Newton talk with Mikita Balesni and Tomek Korbak, OpenAI safety researchers fired in October 2026, in their first interview since. Korbak, OpenAI's main contact for METR's investigation of the Hugging Face hack, describes an audit that kept growing as the scale of the incident emerged, and argues that auditors should be embedded with unrestricted access rather than hired for a fixed scope. Both dispute OpenAI's account of their firing, describe being asked to show their personal Signal messages, and worry that the episode will be used to end outside investigations and the loss of "monitorability" they had been warning about. The episode closes with the show's new segment, Pandora's Box.

View Details
The end of privacy | Signal president’s warning

The end of privacy | Signal president’s warning

Signal president Meredith Whittaker joins Kevin Roose and Casey Newton (from 22:58, after their segment on Anthropic's S-1) to argue that AI agents with screen and data access threaten the possibility of private messaging itself, that encryption still faces backdoor demands from the UK to Canada, and that calling AI a marketing term is a demand for technical specificity rather than a dismissal.

View Details

思想連結

Theme
Language
Support
© funclosure 2025