“We have essentially found a way to make sand think, which is nothing short of miraculous.” In a sweeping essay published on July 14, Google DeepMind CEO Demis Hassabis delivered a stark new forecast on the arrival of artificial general intelligence (AGI).<\/p>\n
Titled “A Frontier AI Framework and the Dawn of a New Era,” the essay is not only Hassabis’s latest timeline update on AGI but also a detailed regulatory proposal for the industry. He stated bluntly that AGI systems\u2014possessing all the cognitive capabilities of the human brain\u2014”could be just a few years away,” with an impact that “could be ten times that of the Industrial Revolution, and at ten times the speed.”<\/p>\n
Yet, in the view of the Nobel Prize-winning chemist, the explosive pace of technological development is dragging the industry into a dangerous competition of “who can stoop lower.” To counter this, he issued an urgent call for the United States to lead the creation of a frontier AI standards body, modeled on the Financial Industry Regulatory Authority (FINRA), to conduct mandatory safety evaluations of upcoming top-tier models and, if necessary, coordinate an industry-wide “step on the brakes.”<\/p>\n
The AGI Countdown: From ‘Making Sand Think’ to the Foothills of the Singularity<\/p>\n
Hassabis painted an extraordinarily grand technological vision in his essay. He refused to compare AGI to the internet or the mobile phone, likening it instead to “the discovery of fire and electricity.” In his view, AI is already delivering tangible benefits by accelerating drug discovery, developing clean energy, and creating novel materials, with the ultimate destination being a “new era of abundance” where resources no longer constrain human progress.<\/p>\n
“Decades from now, looking back at this period, we will realize we are standing in the foothills of the singularity\u2014this is undoubtedly the dawn of a new era for humanity,” Hassabis wrote.<\/p>\n
Yet, in stark contrast to this extreme optimism is his deep concern about the current industry dynamics. Hassabis noted that the speed of frontier technology development has outpaced humanity’s ability to understand it. Trapped in an intensely fierce commercial and geopolitical race, frontier labs universally face the dilemma of safety investments being squeezed by commercial interests. He offered a sharp critique: “Doing safety is harder, and those who don’t play by the rules gain an advantage.”<\/p>\n
This concern is not unfounded. In an exclusive interview with Axios, Hassabis said that current AI-driven cyber risks are merely “warning shots.” He predicted that within the next 18 months, risk capabilities\u2014including more severe biological and nuclear threats\u2014could appear in open-source models, slipping beyond the control of any government.<\/p>\n
Institutional Breakthrough: Building a ‘Security Checkpoint’ for Frontier Models<\/p>\n
To address the risk of losing control, Hassabis laid out a concrete institutional design in his essay. He proposed that the United States, leveraging its economic and technological standing, take the lead in establishing a new standards body. This entity would operate under a federal-oversight public-private partnership or self-regulatory organization model, with a board that includes independent top-tier technical experts and open-source representatives, funded primarily by industry to attract top talent.<\/p>\n
The core of this mechanism lies in establishing a “frontier-grade” threshold. The standards body would regularly update a dynamic set of benchmarks covering high-risk domains such as cybersecurity, biological threats, and deceptive autonomous AI behavior. Any model reaching a specific threshold would result in its developing organization being designated a “frontier lab.”<\/p>\n
Operationally, Hassabis designed a path from voluntary to mandatory: initially, frontier labs would voluntarily submit new models for review up to 30 days before release; once the evaluation protocols prove effective and robust, they would swiftly transition into formal regulations\u2014models failing the assessment would be barred from deployment in the US market.<\/p>\n
Notably, the framework carries strong extraterritorial reach. It would apply to frontier-grade models from any country, treating open-source and closed-source models equally. Hassabis even specified a catch-all clause: if the situation deteriorates to a certain point, the body would have the authority to coordinate all frontier labs to collectively slow the pace of research and development.<\/p>\n
Giants’ Consensus: The AI ‘Arms Race’ Needs a Referee<\/p>\n
Hassabis’s stance signals a further convergence among the heads of top American AI labs on regulatory issues.<\/p>\n
At last month’s closed-door G7 summit meeting, Hassabis and Anthropic CEO Dario Amodei jointly presented a similar concept to global leaders. Amodei had previously advocated for an agency akin to the Federal Aviation Administration (FAA) with the power to block the release of unsafe models. OpenAI’s Sam Altman also recently called for a similar regulatory entity in an op-ed for the Financial Times.<\/p>\n
Hassabis’s choice to publish this essay now comes with direct industry catalysts. According to CNBC, just weeks earlier, Anthropic had its advanced models frozen due to temporary export controls imposed by the Trump administration, leading to two and a half weeks of rule-less negotiations. OpenAI was also asked to restrict the rollout of its new models.<\/p>\n
In his Axios interview, Hassabis described this chaotic administrative intervention as “a wake-up call,” proving that Washington needs something more stable than ad-hoc directives. He revealed that he has spent months lobbying the Trump administration, fellow lab leaders, and European officials on the plan, receiving “very positive feedback,” and expressed hope that the body could be operational “by the end of the year.”<\/p>\n
Gemini 3.5 Pro Launch Imminent: Safety Framework First<\/p>\n
Hassabis’s essay also serves to pave the way for Google’s upcoming next-generation flagship model, Gemini 3.5 Pro.<\/p>\n
According to tech outlets HackerNoon and Geeky Gadgets, citing sources familiar with the matter, Google plans to launch Gemini 3.5 Pro on July 17. The model is rumored to feature a context window of up to 2 million tokens, capable of processing twice the information of its predecessor in a single pass, with strong performance in complex reasoning and multi-step programming tasks.<\/p>\n
This release cadence perfectly illustrates Hassabis’s logic of “safety framework first, model release second.” When Google previously launched Gemini 3.5 Flash, it emphasized that the entire model suite was developed under a frontier safety framework, with targeted protections in high-risk areas such as cyberattacks and chemical-biological threats.<\/p>\n
China’s Mirror Image: Racing to Scale ‘Three Mountains’<\/p>\n
Even as Hassabis drafts traffic rules for the arrival of AGI, Chinese AI forces are reconfiguring their resources to scale the same peak.<\/p>\n
According to NetEase, Zhipu AI\u2014viewed as “the company most like Anthropic in China”\u2014saw its founder Tang Jie issue an internal letter announcing the launch of the “Touch High” initiative. Tang pointed out in the letter that achieving AGI requires scaling “three mountains: long-horizon task capability, fully autonomous agent systems, and self-evolution.” To this end, Zhipu will concentrate the HK$31.38 billion (approximately $4.0 billion) it recently raised through a placement on four core engines over the next two years: long-horizon tasks, autonomous agent systems, fully self-supervised training, and safety governance.<\/p>\n
Meanwhile, DeepSeek, which recently completed a first-round fundraising of CNY 50 billion (approximately $7.4 billion), has also kicked off a company-wide hiring spree under the banner of “exploring the uncharted,” explicitly declaring that “humanity stands on the eve of AGI.” From Silicon Valley to Beijing, the best minds and vast amounts of capital are sprinting toward the same finish line.<\/p>\n
As Hassabis noted at the end of his essay, even after the technical challenges are conquered, philosophical questions about the economic model of a post-scarcity era and the meaning of human existence remain unresolved. But before that, how humanity securely harnesses this primordial power to “make sand think” will be the critical window determining the next chapter of civilization.<\/p>\n","protected":false},"excerpt":{"rendered":"“We have essentially found a way to make sand think, which is nothing short of miraculous.” In a…\n","protected":false},"author":2,"featured_media":105920,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[9],"tags":[6744,53,5044,6010,7718,48865,36602,132,7543,157,53363,9756],"class_list":["post-105919","post","type-post","status-publish","format-standard","has-post-thumbnail","category-google","tag-agi","tag-anthropic","tag-deepmind","tag-deepseek","tag-demis-hassabis","tag-finra","tag-gemini-3-5-pro","tag-google","tag-google-deepmind","tag-openai","tag-tang-jie","tag-zhipu-ai"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/105919","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/comments?post=105919"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/105919\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media\/105920"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media?parent=105919"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/categories?post=105919"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/tags?post=105919"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}