Zuckerberg: AI Should Accelerate, Not Slow Down
Meta founder and CEO Mark Zuckerberg stated in a column for The Wall Street Journal and related comments that "we should accelerate AI development, not limit it," clearly opposing current calls within the industry to slow down AI development.
Zuckerberg specifically mentioned that "the public discussion around AI is not optimistic enough," and that "people spend too much time discussing what could go wrong," stating his goal is to "put personal superintelligence in the hands of everyone"; he also warned that "if we slow down while others continue to advance, it won't put us in a better position."
On specific policy positions, he opposed the proposal to ban Chinese open-source AI models, stating, "I don’t think banning Chinese open-source models is the right approach," advocating that the U.S. should "make its own systems better" to compete rather than relying on restricting competitors.
Financially, Meta's capital expenditure plan for this year is as high as $130 billion to $145 billion for data center construction, currently advancing at "full speed"; the company is also maintaining parallel paths for open-source (LLaMA series continuous iteration) and closed-source (Muse Spark 1.1) models, competing with OpenAI, Google, Anthropic, and xAI in computing power and talent recruitment.
This statement directly confronts the current "slowdown faction" within the industry—Anthropic CEO Dario Amodei recently called for "slowing the pace at the frontier," proposing the introduction of independent safety assessors and coordination of safety standards among major Western labs; OpenAI CEO Sam Altman stated that "slowing down the development pace has been one of the core topics of long-term discussion within OpenAI," and announced the opening of systems to independent assessment agencies; Elon Musk responded that "Dario is right," despite previously criticizing Anthropic as "hating Western civilization" and a "robot cult."
In terms of market mechanisms, the public confrontation between the "accelerators" and "slowdown faction" reveals clear interests at stake—David Sacks, a former White House AI policy official, publicly criticized the "slowdown" proposals from Amodei, Altman, and Musk as "disguised market monopolization under the guise of virtue," stating, "Stop pretending you need to pause antitrust laws to form a cartel," and pointed out that these safety assessment agencies are intertwined with the interests of Anthropic's investors and employees, which objectively burdens startups more than leading firms; notably, the slowdown-calling Anthropic is itself racing towards a trillion-dollar IPO, leading some critics to interpret the "slowdown initiative" as a strategy for leading companies to lock in first-mover advantages, while Zuckerberg's Meta represents the most aggressive capital expenditure and the most resolute public stance in the "acceleration camp."
Source: Public Information
ABAB AI Insight
Zuckerberg's "acceleration theory" continues Meta's consistent path in AI strategy over the past few years—the company previously established a secret "superintelligence lab" to pursue AI capabilities beyond human levels and has continuously increased its investment in the computing power race. This aligns with Meta's historical strategic shifts such as "mobile-first" and "metaverse bets": once a direction is set, it forces the industry into the competitive rhythm it defines through massive capital expenditures.
Financially, Meta's capital expenditure this year is as high as $130 billion to $145 billion, almost entirely directed towards data centers, self-developed chips, and recruiting top research talent, reaching levels comparable to the defense budgets of many countries; this "using scale to crush uncertainty" approach is motivated by Zuckerberg's judgment that if the AI capability race is slowed down by external restrictions, the capital and infrastructure advantages accumulated by Meta over the past few years will be diluted—he has explicitly expressed concerns that "if we slow down while others do not, it is not beneficial for us," which is a typical "first-mover fear of slowing down" mentality.
This public split between "acceleration vs. slowdown" resembles historical patterns in multiple technological races where "incumbents advocate for acceleration, while newcomers or regulators advocate for a pause"—the party with first-mover advantage often prefers to "keep running," while those hoping to change the competitive landscape or fearing being surpassed tend to call for a "stop"; the unique aspect of the current AI industry is that even within the leading camp (Amodei, Altman, Musk), there is a shift towards a "slowdown" stance, while Zuckerberg is almost the only leader of a major lab publicly and firmly opposing the slowdown, positioning Meta as the "last island of accelerationism" in the industry.
Essentially, this is a game about pricing power and competitive landscape, rather than a simple divergence in safety concepts—David Sacks' "cartel" accusation reveals the core of the mechanism: if the "independent safety assessment + industry coordination standards" promoted by Amodei and others are truly implemented, the assessment thresholds and compliance costs will disproportionately burden resource-scarce startups and newcomers, while leading players like Anthropic and OpenAI, who have already established relationships with assessment agencies and are about to complete IPO financing, will find it easier to absorb these costs, effectively locking the industry competitive landscape in favor of the current few leading companies under the guise of "safety"; Zuckerberg's acceleration stance is the only major voice attempting to break this potential lock-in and maintain an open competitive landscape.
Note: Zuckerberg's "acceleration" statement comes from his column and related interviews published in The Wall Street Journal at the end of July, which predates Amodei's recent call for a slowdown; the two are not direct confrontations at the same time point but represent the ongoing positions of the two factions within the industry, hence they are compared and interpreted together.
ABAB News · Cognitive Laws
- Only leaders dare to call for acceleration; laggards always call for slowdown.
- The more detailed the safety standards, the less friendly they are to small players.
- A true cartel never needs to sign contracts; it only needs to say "for safety."