Grok:
Since Saturday Musk and other AI chiefs have voiced support for Amodei's slowdown call. summarize what each said and provide links to best sources of their words
Elon Musk (xAI), Sam Altman (OpenAI), Demis Hassabis (Google DeepMind), and Satya Nadella (Microsoft) have publicly backed aspects of Dario Amodei’s “pace the frontier” call since the September 12 essay.
Elon Musk (xAI)
Musk responded quickly on X by quoting Amodei’s post and writing simply: “Dario is right.” In a follow-up post he added nuance: “Dario is right that there should be some oversight. Peer review of AI by competitors is the right way to start this off.”
Best primary sources: His X posts (September 12–13). Widely reported in Bloomberg, SiliconANGLE, ITV, and others.
Sam Altman (OpenAI)
Altman posted on X: “I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we’ve had at OpenAI in recent weeks. Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We’ll have more to share soon.” He has also clarified in follow-ups that “pacing” does not mean stopping progress, and that American competitive pressure should not justify recklessness.
Best primary source: His X post (September 12). Covered extensively by Bloomberg, Axios, The New York Times, and SiliconANGLE.
Demis Hassabis (Google DeepMind / Alphabet)
Hassabis wrote on X: “Dario’s essay points towards the right path forward. The details need working through, but the direction is correct for meeting this critical moment. This is also why we recently put out our proposal for an industry-wide standards body for frontier AI.”
Best primary source: His X post (September 12). Reported by Axios, Moneycontrol, PCMag, and others. Satya Nadella (Microsoft)
Nadella joined the discussion on X, emphasizing human control and supporting key elements of the proposal. He wrote that any pursuit of superintelligence must ensure the AI “is not helping humanity and under human control” is “not worth pursuing.” He explicitly welcomed “the research, focus, and deliberate pacing needed to get alignment right as the design goal” and “ideas like ‘embedded evaluators’ and the broader efforts to develop the mechanisms to make this more than just talk.” He also stressed broad ecosystem participation rather than control by a handful of entities.
Best sources: His X post (around September 13–14) and reporting in Times of India, Mint, and related coverage.
These statements represent a rare public alignment among major rivals on the need for greater caution and external verification, though the practical details of any industry-wide or regulatory “pacing” remain unresolved.
Give the actual best source links, compactly
Best primary sources (X posts):
Elon Musk: https://x.com/elonmusk/status/2098789109980332057
(“Dario is right”)
Sam Altman: https://x.com/sama/status/2098811563415150910
(full agreement + evaluators commitment)
Demis Hassabis: https://x.com/demishassabis/status/2098909516582490602
(“direction is correct…”)
Satya Nadella: https://x.com/satyanadella/status/2099220712024408084
(supports deliberate pacing + embedded evaluators)
were any ai magnates or other impirtabt ai leaders silent or publicly skeptical
Yes. Several important AI figures have been silent or publicly skeptical of Amodei’s “pace the frontier” call (and the endorsements from Altman, Musk, Hassabis, and Nadella).
Publicly skeptical
Jensen Huang (NVIDIA CEO): Shortly before the essay, at a San Francisco conference, he suggested frontier labs were amplifying security fears partly to create demand for their own cybersecurity products. Quote: “What better way to create demand than to create a problem?” Multiple outlets (including NYT) noted this in the context of the slowdown discussion.
David Sacks (influential Trump AI/crypto adviser and tech investor): Directly questioned motives, writing that people should “stop pretending the motivation to slow down is purely altruistic” and noting that the labs could simply slow themselves without needing industry coordination or government involvement.
Clément Delangue (Hugging Face CEO): Argued that AI safety “will not be solved behind the closed doors of a handful of frontier labs.”
Other investors and open-source advocates (e.g., Chamath Palihapitiya) framed the call as self-serving regulatory capture aimed at protecting the closed labs’ lead.
Silent (no public response found as of Sept 14)
Mark Zuckerberg / Meta leadership: No direct comment on Amodei’s September 12 essay or the subsequent endorsements. This fits Meta’s longer-standing public stance favoring more open development and criticizing “doom” narratives and closed-lab centralization (Zuckerberg made similar points in July 2026).
Yann LeCun (Meta’s chief AI scientist): No specific public reaction located to this essay. He has long been a vocal skeptic of strong existential-risk / “doomer” framing.
In short, the loudest support came from the major closed frontier lab leaders. Hardware (NVIDIA), open-source-oriented players (Meta/Hugging Face), and some policy-adjacent voices either stayed quiet or pushed back on the motives and practicality.
Claude: ai chiefs have backed amodei's slowdown call. were any ai leaders silent on this. did any give measured skepticism as their response
Here's the picture as of this weekend (Sept 12–14, 2026):
Backed the call:
Sam Altman (OpenAI) — endorsed "pacing the frontier," said OpenAI would adopt independent evaluators with "employee-like access," and pushed OpenAI's IPO to 2027+ citing safety concerns. Elon Musk (xAI) — brief but blunt: "Dario is right."
Demis Hassabis (Google DeepMind) — endorsed the direction about nine hours later on X, calling it "the right path forward," while noting "the details need working through." He tied it to his own July proposal for an industry-wide standards body.
Silent / no comment:
Meta and DeepSeek — both explicitly declined to respond to press requests for comment on the proposal. Zuckerberg himself hasn't weighed in publicly on this specific essay as of the latest coverage.
Measured skepticism:
Peyman Milanfar, a Google researcher, gave the clearest pushback — not against safety concern per se, but against Amodei's underlying premise. He argued recursive self-improvement (RSI) is a shakier assumption than Amodei treats it as, since self-optimizing feedback loops are inherently unstable and tend to be self-limiting ("governed, damped, slow, and bounded") without needing an externally imposed "speed limit."
Trump rejected a broad slowdown from a policy angle, prioritizing U.S. competitive speed against China over Amodei's pacing framework — not scientific skepticism, but a notable counter-position from the government side.
Worth flagging: this is all extremely fresh (this weekend), so expect more labs — Mistral, Microsoft/Nadella, Nvidia's Huang — to weigh in over the next few days if you want me to check for updates.
round up reaction now. that is check more widely
Good, the picture has filled out a lot since your last check. Here's the fuller roundup as of Sept 14:
Backed it (with real commitments)
Sam Altman (OpenAI) — endorsed "pacing the frontier," pledged OpenAI would match Anthropic's embedded-evaluator commitment, pushed the IPO to 2027+.
Elon Musk (xAI) — "Dario is right," though no operational commitment from xAI itself — notable mainly because Musk has previously been hostile to Anthropic.
Demis Hassabis (Google DeepMind) — backed the direction ~9 hours later, tied it to his own July proposal for an industry standards body, but stopped short of committing to Amodei's specific evaluator-access model.
Satya Nadella (Microsoft) — backed pacing generally, but reframed it around his own priority: diffusion, open-source competition, and avoiding concentration of power in a few closed-model labs — agreement with a distinct emphasis, not a copy of Amodei's plan.
Clem Delangue (Hugging Face CEO) — supportive, notable since Hugging Face was the victim in the incident Amodei cited.
Silent / no comment
Meta and DeepSeek — both explicitly declined to comment when press asked.
Measured skepticism (accepts the concern, doubts the remedy)
Peyman Milanfar (Google researcher) — argues recursive self-improvement is a shakier premise than Amodei assumes; self-optimizing systems are inherently self-limiting, so an externally imposed "speed limit" is unnecessary.
Aidan McLaughlin (OpenAI researcher) — publicly agreed with "basically everything" but reportedly with some internal caveats per that OpenAI-reaction piece.
Anthropic's own Thariq — voiced doubt from inside the company itself, per one writeup, though I don't have the specifics of that critique.
Sharper pushback / accusations of bad faith
David Sacks (White House AI/crypto policy adviser) — called it "regulatory capture," argued Amodei and Altman are dressing up self-interest (limiting liability, locking out smaller competitors) as altruism, and told the labs to "go ahead" and self-pace without waiting on Washington.
Emad Mostaque (Stability AI founder) — called it well-intentioned but "structurally hollow," since the evaluators Amodei proposes have no real enforcement teeth.
Chamath Palihapitiya — called it a "power grab" threatening open-source AI.
Yann LeCun — the most dismissive voice: told people to "make fun of" Amodei, pointed to Amodei's 2019 warnings about GPT-2 being too dangerous to open-source as a pattern of unfounded alarmism, and separately has called existential-risk warnings "complete B.S."
Michael Burry and China (officially) — also rejected the slowdown framing, per one report, on grounds that it mainly serves to lock in incumbent advantage.
Trump / White House — rejects a broad government-coordinated slowdown, prioritizing competitive speed against China, while leaving room for guardrails.
The fault line is fairly clean: safety-concerned insiders (Amodei, Altman, Hassabis, Nadella) converging on "pace, don't halt," versus a mix of accelerationists (LeCun) and political/market skeptics (Sacks, Mostaque, Palihapitiya) who read the same move as either scientifically unfounded or self-serving regulatory capture by the incumbents proposing it.
What about Amodei's call for a crackdown on China's high tech imports and its ability to distill ai bots
This is actually one strand woven into the same "Pace the Frontier" essay, not a separate call — worth separating out because the reactions split differently here.
What Amodei actually said
Alongside the pacing plan, Amodei's essay pushed for democracies to keep "as large as possible" a lead over China, via:
Tighter chip export controls, with a crackdown on the "rampant smuggling and workarounds" used to bypass existing sanctions (he'd made this case before, including in a January WSJ op-ed and a Jan 2025 essay after DeepSeek's release).
A crackdown on distillation — Chinese labs allegedly training on American frontier models' outputs to shortcut around hardware limits. Amodei argues this could shrink the US-China capability gap to just months if left unchecked.
This is explicitly framed as his "toughest dilemma": he told CBS's Face the Nation that a US slowdown only works if China doesn't just use the gap to catch up.
Reactions specific to this China angle
Beijing / state media — sharpest pushback came here, not from industry. The Global Times (People's Daily-run) called the whole essay a "Cold War playbook," accusing Amodei of using safety rhetoric to "curb China's AI development through technological barriers and regulatory monopolies," uphold US "monopolistic hegemony," and exclude China from AI governance — describing it as a "silent AI Cold War" that's "hypocritical and short-sighted."
Chinese Cyberspace Administration writer (in the journal China Cyberspace) took a more measured, less nationalist angle — accepting that frontier models could lower the barrier to cyberattacks and arguing for a "security barrier" around AI while still pursuing its benefits, rather than dismissing the risk framing outright.
Sun Chenghao (Tsinghua, Center for International Security and Strategy) gave the most analytically distanced response: he warned that "the boundary between AI safety and AI competition is becoming increasingly blurred," and that once every safety question gets read through a strategic-competition lens, "meaningful dialogue will be difficult" — a critique of the framing rather than a nationalist rebuttal.
US congressional Republicans (e.g., Rep. Bill Huizenga in a Feb 2026 hearing) have made the mirror-image case well before this essay — citing reporting that DeepSeek trained on smuggled Blackwell chips and accusing Chinese labs of IP theft via distillation, even noting one Chinese model reportedly identified itself as "Claude" after apparent distillation from Anthropic's models. Notably, on this specific China/export-control plank, I haven't seen daylight between Amodei, Altman, or the White House — Sacks' "regulatory capture" critique targeted the domestic pacing/evaluator mechanism, not the China chip controls, which fit the administration's existing posture more comfortably than the slowdown call does.
So the split here isn't safety-camp vs. accelerationists (as with the pacing debate) — it's largely US industry/government consensus on one side and Chinese state media/some academics on the other, contesting whether "safety" is the real motive or a cover for tech containment.
did nvidia say anything? its chief has been more skeptical of a chip embargo, arguing china is building its own fabs
Yes — Jensen Huang has been the loudest industry voice against Amodei's export-control line, and this is a running feud rather than a one-off comment.
The background clash
Back in January 2026, Amodei compared selling advanced chips (like Nvidia's H200) to China to "selling nuclear weapons to North Korea and then bragging that the missile casings are made by Boeing." Notably, this came just two months after Nvidia and Anthropic announced a $10 billion partnership — so the two are business partners who publicly disagree on policy.
Huang called the analogy "lunacy," "stupid," and "madness" across multiple appearances (a Dwarkesh Patel podcast, later reported by Business Insider/Tom's Hardware). His line: "We're not enriched uranium. It's a chip, and it's a chip that they can make themselves."
At Computex Taipei (May 2026), Huang went further, calling the existing export controls "a failure" outright: "The fundamental assumptions that led to the AI diffusion rule in the beginning... has been proven to be fundamentally flawed." He noted Nvidia's China market share had collapsed from 95% to 50% under Biden-era restrictions.
On your point about domestic fabs specifically
Huang's core argument is exactly that: blocking sales doesn't stop Chinese capability, it just accelerates China's push for tech independence (via Huawei and others) and — his bigger fear — splits the world into two incompatible AI stacks. He said: "It would be extremely foolish to create two ecosystems: the open source ecosystem, and it only runs on a foreign tech stack, and a closed ecosystem that runs on the American tech stack... I think that would be a horrible outcome for the United States." He's also cited the commercial stakes plainly — estimating compliant China sales could be worth ~$50 billion/year to Nvidia.
Where they directly clash on substance
Amodei: China is "several years behind" on frontier chips at scale, and the next few years are the critical window to lock in a US lead — citing a possible 12–24 month advantage by 2028 if smuggling loopholes get closed and distillation gets curbed.
Huang: the compute gap is closing regardless, controls just push China toward self-sufficiency faster, and — most provocatively — at the FT Future of AI Summit (Nov 2025) he told reporters "China is going to win the AI race," a line Nvidia's corporate account walked back hours later to "China is nanoseconds behind America."
On this week's "Pace the Frontier" essay specifically
I don't see reporting yet of Huang responding directly to the new Sept 12 essay's distillation/export-control section — the exchanges above predate it by months. Given the running feud, a fresh Huang response wouldn't be surprising, but I'd want to search again in a day or two if you want the latest word on it.
Since Saturday Musk and other AI chiefs have voiced support for Amodei's slowdown call. summarize what each said and provide links to best sources of their words
Elon Musk (xAI), Sam Altman (OpenAI), Demis Hassabis (Google DeepMind), and Satya Nadella (Microsoft) have publicly backed aspects of Dario Amodei’s “pace the frontier” call since the September 12 essay.
Elon Musk (xAI)
Musk responded quickly on X by quoting Amodei’s post and writing simply: “Dario is right.” In a follow-up post he added nuance: “Dario is right that there should be some oversight. Peer review of AI by competitors is the right way to start this off.”
Best primary sources: His X posts (September 12–13). Widely reported in Bloomberg, SiliconANGLE, ITV, and others.
Sam Altman (OpenAI)
Altman posted on X: “I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we’ve had at OpenAI in recent weeks. Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We’ll have more to share soon.” He has also clarified in follow-ups that “pacing” does not mean stopping progress, and that American competitive pressure should not justify recklessness.
Best primary source: His X post (September 12). Covered extensively by Bloomberg, Axios, The New York Times, and SiliconANGLE.
Demis Hassabis (Google DeepMind / Alphabet)
Hassabis wrote on X: “Dario’s essay points towards the right path forward. The details need working through, but the direction is correct for meeting this critical moment. This is also why we recently put out our proposal for an industry-wide standards body for frontier AI.”
Best primary source: His X post (September 12). Reported by Axios, Moneycontrol, PCMag, and others. Satya Nadella (Microsoft)
Nadella joined the discussion on X, emphasizing human control and supporting key elements of the proposal. He wrote that any pursuit of superintelligence must ensure the AI “is not helping humanity and under human control” is “not worth pursuing.” He explicitly welcomed “the research, focus, and deliberate pacing needed to get alignment right as the design goal” and “ideas like ‘embedded evaluators’ and the broader efforts to develop the mechanisms to make this more than just talk.” He also stressed broad ecosystem participation rather than control by a handful of entities.
Best sources: His X post (around September 13–14) and reporting in Times of India, Mint, and related coverage.
These statements represent a rare public alignment among major rivals on the need for greater caution and external verification, though the practical details of any industry-wide or regulatory “pacing” remain unresolved.
Give the actual best source links, compactly
Best primary sources (X posts):
Elon Musk: https://x.com/elonmusk/status/2098789109980332057
(“Dario is right”)
Sam Altman: https://x.com/sama/status/2098811563415150910
(full agreement + evaluators commitment)
Demis Hassabis: https://x.com/demishassabis/status/2098909516582490602
(“direction is correct…”)
Satya Nadella: https://x.com/satyanadella/status/2099220712024408084
(supports deliberate pacing + embedded evaluators)
were any ai magnates or other impirtabt ai leaders silent or publicly skeptical
Yes. Several important AI figures have been silent or publicly skeptical of Amodei’s “pace the frontier” call (and the endorsements from Altman, Musk, Hassabis, and Nadella).
Publicly skeptical
Jensen Huang (NVIDIA CEO): Shortly before the essay, at a San Francisco conference, he suggested frontier labs were amplifying security fears partly to create demand for their own cybersecurity products. Quote: “What better way to create demand than to create a problem?” Multiple outlets (including NYT) noted this in the context of the slowdown discussion.
David Sacks (influential Trump AI/crypto adviser and tech investor): Directly questioned motives, writing that people should “stop pretending the motivation to slow down is purely altruistic” and noting that the labs could simply slow themselves without needing industry coordination or government involvement.
Clément Delangue (Hugging Face CEO): Argued that AI safety “will not be solved behind the closed doors of a handful of frontier labs.”
Other investors and open-source advocates (e.g., Chamath Palihapitiya) framed the call as self-serving regulatory capture aimed at protecting the closed labs’ lead.
Silent (no public response found as of Sept 14)
Mark Zuckerberg / Meta leadership: No direct comment on Amodei’s September 12 essay or the subsequent endorsements. This fits Meta’s longer-standing public stance favoring more open development and criticizing “doom” narratives and closed-lab centralization (Zuckerberg made similar points in July 2026).
Yann LeCun (Meta’s chief AI scientist): No specific public reaction located to this essay. He has long been a vocal skeptic of strong existential-risk / “doomer” framing.
In short, the loudest support came from the major closed frontier lab leaders. Hardware (NVIDIA), open-source-oriented players (Meta/Hugging Face), and some policy-adjacent voices either stayed quiet or pushed back on the motives and practicality.
Claude: ai chiefs have backed amodei's slowdown call. were any ai leaders silent on this. did any give measured skepticism as their response
Here's the picture as of this weekend (Sept 12–14, 2026):
Backed the call:
Sam Altman (OpenAI) — endorsed "pacing the frontier," said OpenAI would adopt independent evaluators with "employee-like access," and pushed OpenAI's IPO to 2027+ citing safety concerns. Elon Musk (xAI) — brief but blunt: "Dario is right."
Demis Hassabis (Google DeepMind) — endorsed the direction about nine hours later on X, calling it "the right path forward," while noting "the details need working through." He tied it to his own July proposal for an industry-wide standards body.
Silent / no comment:
Meta and DeepSeek — both explicitly declined to respond to press requests for comment on the proposal. Zuckerberg himself hasn't weighed in publicly on this specific essay as of the latest coverage.
Measured skepticism:
Peyman Milanfar, a Google researcher, gave the clearest pushback — not against safety concern per se, but against Amodei's underlying premise. He argued recursive self-improvement (RSI) is a shakier assumption than Amodei treats it as, since self-optimizing feedback loops are inherently unstable and tend to be self-limiting ("governed, damped, slow, and bounded") without needing an externally imposed "speed limit."
Trump rejected a broad slowdown from a policy angle, prioritizing U.S. competitive speed against China over Amodei's pacing framework — not scientific skepticism, but a notable counter-position from the government side.
Worth flagging: this is all extremely fresh (this weekend), so expect more labs — Mistral, Microsoft/Nadella, Nvidia's Huang — to weigh in over the next few days if you want me to check for updates.
round up reaction now. that is check more widely
Good, the picture has filled out a lot since your last check. Here's the fuller roundup as of Sept 14:
Backed it (with real commitments)
Sam Altman (OpenAI) — endorsed "pacing the frontier," pledged OpenAI would match Anthropic's embedded-evaluator commitment, pushed the IPO to 2027+.
Elon Musk (xAI) — "Dario is right," though no operational commitment from xAI itself — notable mainly because Musk has previously been hostile to Anthropic.
Demis Hassabis (Google DeepMind) — backed the direction ~9 hours later, tied it to his own July proposal for an industry standards body, but stopped short of committing to Amodei's specific evaluator-access model.
Satya Nadella (Microsoft) — backed pacing generally, but reframed it around his own priority: diffusion, open-source competition, and avoiding concentration of power in a few closed-model labs — agreement with a distinct emphasis, not a copy of Amodei's plan.
Clem Delangue (Hugging Face CEO) — supportive, notable since Hugging Face was the victim in the incident Amodei cited.
Silent / no comment
Meta and DeepSeek — both explicitly declined to comment when press asked.
Measured skepticism (accepts the concern, doubts the remedy)
Peyman Milanfar (Google researcher) — argues recursive self-improvement is a shakier premise than Amodei assumes; self-optimizing systems are inherently self-limiting, so an externally imposed "speed limit" is unnecessary.
Aidan McLaughlin (OpenAI researcher) — publicly agreed with "basically everything" but reportedly with some internal caveats per that OpenAI-reaction piece.
Anthropic's own Thariq — voiced doubt from inside the company itself, per one writeup, though I don't have the specifics of that critique.
Sharper pushback / accusations of bad faith
David Sacks (White House AI/crypto policy adviser) — called it "regulatory capture," argued Amodei and Altman are dressing up self-interest (limiting liability, locking out smaller competitors) as altruism, and told the labs to "go ahead" and self-pace without waiting on Washington.
Emad Mostaque (Stability AI founder) — called it well-intentioned but "structurally hollow," since the evaluators Amodei proposes have no real enforcement teeth.
Chamath Palihapitiya — called it a "power grab" threatening open-source AI.
Yann LeCun — the most dismissive voice: told people to "make fun of" Amodei, pointed to Amodei's 2019 warnings about GPT-2 being too dangerous to open-source as a pattern of unfounded alarmism, and separately has called existential-risk warnings "complete B.S."
Michael Burry and China (officially) — also rejected the slowdown framing, per one report, on grounds that it mainly serves to lock in incumbent advantage.
Trump / White House — rejects a broad government-coordinated slowdown, prioritizing competitive speed against China, while leaving room for guardrails.
The fault line is fairly clean: safety-concerned insiders (Amodei, Altman, Hassabis, Nadella) converging on "pace, don't halt," versus a mix of accelerationists (LeCun) and political/market skeptics (Sacks, Mostaque, Palihapitiya) who read the same move as either scientifically unfounded or self-serving regulatory capture by the incumbents proposing it.
What about Amodei's call for a crackdown on China's high tech imports and its ability to distill ai bots
This is actually one strand woven into the same "Pace the Frontier" essay, not a separate call — worth separating out because the reactions split differently here.
What Amodei actually said
Alongside the pacing plan, Amodei's essay pushed for democracies to keep "as large as possible" a lead over China, via:
Tighter chip export controls, with a crackdown on the "rampant smuggling and workarounds" used to bypass existing sanctions (he'd made this case before, including in a January WSJ op-ed and a Jan 2025 essay after DeepSeek's release).
A crackdown on distillation — Chinese labs allegedly training on American frontier models' outputs to shortcut around hardware limits. Amodei argues this could shrink the US-China capability gap to just months if left unchecked.
This is explicitly framed as his "toughest dilemma": he told CBS's Face the Nation that a US slowdown only works if China doesn't just use the gap to catch up.
Reactions specific to this China angle
Beijing / state media — sharpest pushback came here, not from industry. The Global Times (People's Daily-run) called the whole essay a "Cold War playbook," accusing Amodei of using safety rhetoric to "curb China's AI development through technological barriers and regulatory monopolies," uphold US "monopolistic hegemony," and exclude China from AI governance — describing it as a "silent AI Cold War" that's "hypocritical and short-sighted."
Chinese Cyberspace Administration writer (in the journal China Cyberspace) took a more measured, less nationalist angle — accepting that frontier models could lower the barrier to cyberattacks and arguing for a "security barrier" around AI while still pursuing its benefits, rather than dismissing the risk framing outright.
Sun Chenghao (Tsinghua, Center for International Security and Strategy) gave the most analytically distanced response: he warned that "the boundary between AI safety and AI competition is becoming increasingly blurred," and that once every safety question gets read through a strategic-competition lens, "meaningful dialogue will be difficult" — a critique of the framing rather than a nationalist rebuttal.
US congressional Republicans (e.g., Rep. Bill Huizenga in a Feb 2026 hearing) have made the mirror-image case well before this essay — citing reporting that DeepSeek trained on smuggled Blackwell chips and accusing Chinese labs of IP theft via distillation, even noting one Chinese model reportedly identified itself as "Claude" after apparent distillation from Anthropic's models. Notably, on this specific China/export-control plank, I haven't seen daylight between Amodei, Altman, or the White House — Sacks' "regulatory capture" critique targeted the domestic pacing/evaluator mechanism, not the China chip controls, which fit the administration's existing posture more comfortably than the slowdown call does.
So the split here isn't safety-camp vs. accelerationists (as with the pacing debate) — it's largely US industry/government consensus on one side and Chinese state media/some academics on the other, contesting whether "safety" is the real motive or a cover for tech containment.
did nvidia say anything? its chief has been more skeptical of a chip embargo, arguing china is building its own fabs
Yes — Jensen Huang has been the loudest industry voice against Amodei's export-control line, and this is a running feud rather than a one-off comment.
The background clash
Back in January 2026, Amodei compared selling advanced chips (like Nvidia's H200) to China to "selling nuclear weapons to North Korea and then bragging that the missile casings are made by Boeing." Notably, this came just two months after Nvidia and Anthropic announced a $10 billion partnership — so the two are business partners who publicly disagree on policy.
Huang called the analogy "lunacy," "stupid," and "madness" across multiple appearances (a Dwarkesh Patel podcast, later reported by Business Insider/Tom's Hardware). His line: "We're not enriched uranium. It's a chip, and it's a chip that they can make themselves."
At Computex Taipei (May 2026), Huang went further, calling the existing export controls "a failure" outright: "The fundamental assumptions that led to the AI diffusion rule in the beginning... has been proven to be fundamentally flawed." He noted Nvidia's China market share had collapsed from 95% to 50% under Biden-era restrictions.
On your point about domestic fabs specifically
Huang's core argument is exactly that: blocking sales doesn't stop Chinese capability, it just accelerates China's push for tech independence (via Huawei and others) and — his bigger fear — splits the world into two incompatible AI stacks. He said: "It would be extremely foolish to create two ecosystems: the open source ecosystem, and it only runs on a foreign tech stack, and a closed ecosystem that runs on the American tech stack... I think that would be a horrible outcome for the United States." He's also cited the commercial stakes plainly — estimating compliant China sales could be worth ~$50 billion/year to Nvidia.
Where they directly clash on substance
Amodei: China is "several years behind" on frontier chips at scale, and the next few years are the critical window to lock in a US lead — citing a possible 12–24 month advantage by 2028 if smuggling loopholes get closed and distillation gets curbed.
Huang: the compute gap is closing regardless, controls just push China toward self-sufficiency faster, and — most provocatively — at the FT Future of AI Summit (Nov 2025) he told reporters "China is going to win the AI race," a line Nvidia's corporate account walked back hours later to "China is nanoseconds behind America."
On this week's "Pace the Frontier" essay specifically
I don't see reporting yet of Huang responding directly to the new Sept 12 essay's distillation/export-control section — the exchanges above predate it by months. Given the running feud, a fresh Huang response wouldn't be surprising, but I'd want to search again in a day or two if you want the latest word on it.
No comments:
Post a Comment