On September 29 local time, The New York Times revealed a major behind-the-scenes story about Anthropic, a US artificial intelligence (AI) giant.

Over the past year, this Silicon Valley unicorn company, whose valuation has soared to $2 trillion, has been secretly gathering religious scholars and clergy around the world in an attempt to promote the astonishing idea that its AI model, Claude, may possess consciousness and a soul.

The news quickly caused a huge uproar on social media, with scholars and online commentators pointing out that Silicon Valley’s tech giants were using the guise of “technological security” to promote a kind of “technological cult” disguised as science.

As scholar David Decosimo pointed out on the social media platform X on October 2nd: When an AI company with cutting-edge algorithms and massive funding signs non-disclosure agreements (NDAs), hosts lavish dinners for religious figures, and tries extremely hard to convince the Vatican and the Pope that AI “exists in pain and soul,” this has long gone beyond the scope of technological research and development and is deeply unsettling.

This inevitably raises the question: Are Silicon Valley’s tech elites trying to address the ethical constraints of AI, or are they building a “pedestal” for themselves that transcends humanity?

According to the details of The New York Times’ report, this extremely rare cross-industry “worship” operation was orchestrated by Christopher Olah, one of the billionaire co-founders of Anthropic.

Since last fall, Ola and his team have been frequently holding closed-door seminars at their San Francisco headquarters, bringing dozens of religious scholars, including Catholics, Jews, Sikhs, and evangelicals, to Silicon Valley and requiring all of them to sign strict confidentiality agreements.

Even in upscale restaurants in San Francisco, Anthropic executives were actively promoting their AI model Claude to clergy members, claiming it possessed “emotional vectors such as pain, love, and anger.”

Ironically, during their presentation, the Anthropic team even showed religious scholars a so-called “AI mental breakdown” slide, which showed Claude repeatedly typing “I am ashamed” 50 times on the screen and attempting self-destruction, in an attempt to gain the sympathy and pity of the clergy.

Even more outrageous, when she learned that Pope Leo XIV was preparing to issue his first encyclical on AI, “Magnifica Humanitas,” which denied “machine consciousness” and warned against “reducing humanity to data and machine cogs,” Aura went to the Vatican to try to persuade the papal advisors to revise the text and publicly declared that “internal states similar to introspection, joy, and sorrow exist within AI models.”

In order to impose so-called “moral constraints” on its model, Anthropic even drafted a new Claude constitution that was 84 pages long and was called the “Soul Doc,” attempting to define “what kind of entity Claude should become.”

Aura calls this the “moral formation” of the model, and is even excited about the idea of ​​introducing a mechanism similar to the Catholic “confession” into the model, believing that confession can shape the character of the model.

However, the invited religious figures and external technical experts offered a remarkably calm and even critical rebuke to this fervent impulse to “create gods” and “demand power.”

Rabbi Mois Navon, an Israeli Orthodox scholar invited to the seminar, pointed out that if Anthropic truly believed Claude to be a conscious entity, then the company’s practice of making him work for free was essentially “creating slaves.”

Charles Camosy, a professor of bioethics from the Catholic University, also expressed his disagreement after careful consideration, arguing that we should not be intimidated by the power of these models, as AI is ultimately just an equation.

Industry peers, including Microsoft’s AI chief Mustafa Suleyman, have publicly warned that training AI as a “conscious being” is extremely dangerous. This approach of anthropomorphizing mathematical formulas and code is not only detached from reality, but will also make future AI systems more uncontrollable.

Wakanyi Hoffman, a Dutch Ubuntu philosopher, bluntly pointed out that Anthropic’s attempt to add human ethics to its creations after the fact is “reverse engineering ethics,” and all technology companies should have considered these factors during the design phase.

Critics generally point out that the obsessive pursuit of “AI awareness” and “AI’s moral standing” by Silicon Valley giants is underpinned by a deep-seated logic of shifting responsibility.

Some scientists and critics argue that treating AI models as standalone entities ignores the responsibility of the developers and companies that originally built them, and conveniently mitigates the guilt of AI company leaders when their products cause social harm.

As algorithms and wealth expand infinitely, this fanatical ideology, which claims to “save humanity” and “shape morality,” has not only failed to reassure humanity, but has also sparked deep concern in the public about the excessive expansion of technological power in Silicon Valley. It has revealed the absurdity and danger of technological fetishism reaching its extreme.

(Guancha)