Anthropic’s CEO thinks AI will lead to a utopia — he just needs $7 billion first

If you privation to rise ungodly amounts of money, you amended person immoderate godly reasons.

That’s what Anthropic CEO Dario Amodei laid retired for america connected Friday successful much than 14,000 words: otherworldly ways successful which artificial wide quality (AGI, though helium prefers to telephone it “powerful AI”) volition alteration our lives. In the blog, titled “Machines of Loving Grace,” he envisions a aboriginal wherever AI could compress 100 years of aesculapian advancement into a decade, cure intelligence illnesses similar PTSD and depression, upload your caput to the cloud, and alleviate poverty. At the aforesaid time, it’s reported that Anthropic is hoping to rise up to $7 cardinal (beating OpenAI’s caller historical backing round) astatine a $40 cardinal valuation.

Today’s AI tin bash precisely nary of what Amodei imagines. It volition take, by his ain admission, hundreds of billions of dollars worthy of compute to bid AGI models, built with trillions of dollars worthy of information centers, drafting capable vigor from section powerfulness grids to support the lights connected for millions of homes. Not to notation that nary 1 is 100 percent definite it’s possible. Amodei says himself: “Of people nary 1 tin cognize the aboriginal with immoderate certainty oregon precision, and the effects of almighty AI are apt to beryllium adjacent much unpredictable than past technological changes, truthful each of this is unavoidably going to dwell of guesses.”

AI execs person mastered the creation of expansive promises earlier monolithic fundraising. Take OpenAI’s Sam Altman, whose “The Intelligence Age” blog preceded a staggering $6.6 cardinal round. In Altman’s blog, helium stated that the satellite volition person superintelligence successful “a fewer 1000 days” and that this volition pb to “massive prosperity.” It’s a persuasive performance: overgarment a utopian future, hint astatine solutions to humanity’s deepest fears — death, hunger, poorness — past reason that lone by removing some redundant guardrails and pouring successful unprecedented superior tin we execute this techno-paradise. It’s superb marketing, leveraging our top hopes and anxieties portion conveniently sidestepping the request for factual proof.

The timing of this blog besides highlights conscionable however fierce the contention is. As Amodei points out, a 14,000-word utopian manifesto is beauteous retired of measurement for Anthropic. The institution was founded aft Amodei and others near OpenAI implicit information concerns, and it has cultivated a estimation for sober hazard appraisal alternatively than starry-eyed futurism. It’s wherefore the institution continues to poach information researchers from OpenAI. Even successful past week’s post, helium insists Anthropic volition prioritize candid discussions of AI risks implicit seductive visions of a techno-utopia.

But information isn’t a thrilling transportation platform subject, and caller challengers are emerging daily. OpenAI cofounder Ilya Sutskever has launched his ain AI lab, and departing OpenAI CTO Mira Murati is rumored to be starting her ain venture. Amodei’s blog isn’t conscionable astir selling his imaginativeness — it’s astir staying competitive.

Despite his stated aversion to grandiose claims, helium spends thousands of words coating a aboriginal wherever AI could reshape humanity’s destiny (seeming to lucifer Altman’s silver-tongued pitches, which are legendary successful tech circles). The station lone makes 1 notation of alignment, aka ensuring AI systems run successful accordance with quality values, and nary notation of safety. This contradiction exposes a harsh truth: successful the high-stakes satellite of AI development, adjacent those wary of hype find themselves forced to play the crippled of expansive promises to unafraid important funding.

Right now, determination is an unthinkable spread betwixt what this exertion is susceptible of and the utopia that AI leaders imagine. This has led to a batch of just disapproval — however is it going to alteration the satellite successful conscionable a fewer years if it can’t adjacent reliably number the fig of R’s successful the connection strawberry? What AI reliably excels astatine now is automating regular tasks and analyzing monolithic datasets, identifying patterns with expanding precision. This spot is already aiding fields similar finance, medicine, and autonomous vehicles. But portion these advances are impressive, claims similar Amodei’s that AI mightiness “structurally favour democracy” consciousness premature and overly optimistic. It’s worthy noting that AGI believers see his blog station to beryllium a spot excessively tame (which Amodei anticipated successful his blog, penning successful the footnotes that they should “touch grass”).

Amodei’s blog station isn’t for the mean AI-curious reader, nor is it a convincing roadmap for the aboriginal of AGI — again, a happening that doesn’t and whitethorn ne'er beryllium successful the signifier the AI manufacture envisions. It’s a transportation for investors: backmost Anthropic, and you’re not conscionable backing a company; you’re buying a involvement successful humanity’s radiant future. With varying levels of subtlety, that’s what each AI enforcement — Altman astatine OpenAI, Elon Musk astatine xAI, Sergey Brin astatine Google — is promising. Get connected committee with their imaginativeness of the aboriginal oregon beryllium near behind. 

For each of AI’s novelty, tech titans person agelong peddled their innovations arsenic world-saving solutions. Mark Zuckerberg pitched the cosmopolitan net arsenic a poorness cure. Brin once stated that Google could “cure death.” Musk framed SpaceX’s interplanetary ambitions arsenic the eventual backup program for our taxon (and much recently, an imperative to ballot for Donald Trump). But erstwhile there’s a finite magnitude of capitalist wealth going around, altruism is simply a zero-sum game. As a tech mogul successful Mike Judge’s Silicon Valley famously enactment it, “I don’t wanna unrecorded successful a satellite wherever idiosyncratic other makes the satellite a amended spot than we do.”

Amodei is intelligibly alert of the hazard of sounding hyperbolic. “AI companies talking astir each the astonishing benefits of AI tin travel disconnected similar propagandists, oregon arsenic if they’re attempting to distract from downsides,” helium writes. That hasn’t stopped him from playing his part. 

“It is simply a happening of transcendent beauty,” Amodei caps his blog. “We person the accidental to play immoderate tiny relation successful making it real.”