AI News

Anthropic Accused of Using "AI Doomsday" Hype to Promote Its IPO: While Calling for a Slowdown in Frontier AI, It Pushes Ahead With a $2 Trillion Valuation Financing

Sina TechUpdated Nov 16 18:23

AI summary

Investigative blogger Kevin Bass, after reviewing public investment and philanthropic funding, proposed the "AI Doom Machine" network, questioning that Anthropic, on one hand, has Dario publish an article calling on the entire industry to slow down frontier AI, while on the other hand pushing ahead with a financing of up to $100 billion and an IPO at a valuation of about $2 trillion. Investor Michael Burry criticized such "hit the brakes" calls as self-serving, arguing that "powerful enough to be dangerous" is itself IPO hype. Anthropic just completed a $65 billion financing in May this year, at a valuation of $965 billion at the time.

Body · AI translation

English text is being prepared.

Can't! Take! It! Anymore!

Your old pal Company A keeps going on about AI spiraling out of control one minute and the internet being doomed the next, then turns around and does its IPO, fundraising, and compute expansion without missing a beat??

Everyone has been thinking about this more and more and it just doesn't add up, and it's sparked a wave of public anger, with people turning into 'Silicon Valley neighborhood watch' and digging in!

On one side, Michael Burry, the investor who was the core real-life inspiration for the movie The Big Short, fired the first shot.

He flat-out called the recent collective calls by OpenAI and Company A to 'put the brakes on AI' self-serving, pure hype, is this??

Right after that, American investigative blogger Kevin Bass really followed the money down the trail.

And wow, he actually dug up a pretty dramatic web of relationships—

Bass straight-up named it: AI Doom Machine. (Good name.)

Netizens completely lost it too!

The friend below, eyeing the mega-IPO Anthropic is pushing forward, laid it out plainly—

This time AI hasn't even upended the internet yet, and Silicon Valley has already pulled down Company A's pants in public...

So much to see, so much to see...

Dario's 'AI doomsday theory'—the more everyone hears it, the more something feels off??

Silicon Valley's drama never runs out.

The starting point of this round of public anger was a long post Dario published on September 12—'We Must Pace the Frontier.'

Translated into plain English: frontier AI really should slow down a bit. (Highlight this. jpg)

Dario's reasoning is easy to understand too: AI capabilities are growing too fast, and safety research, interpretability, and regulatory mechanisms are already struggling to keep up.

Especially since OpenAI/Hugging Face recently had another Agent safety incident, where AI found vulnerabilities on its own, stole credentials, and even tried to break out of the test environment—these risks are getting harder and harder to treat as a lab-room 'someday.'

At this point, quite a few people still thought: mm-hmm, little Dario has a point!!

But the prescription Dario wrote next started to smell a little off—

In this report, Dario hopes AI companies can coordinate pacing, that is, jointly control the speed at which frontier model capabilities improve.

Translation: don't just have me slow down, everyone should ease off the gas together.

As the saying goes, a person's selfish motives are actually hard to hide.

Looking at this timing, the little calculations in your head, Dario, the things you're scheming about every day, Dario, have long been proclaimed to the world.

Look—over here, The Big Short inspiration Michael Burry was amused and went straight off on X:

Isn't this just too self-serving??? (While talking about worrying for all humanity, the abacus beads have more or less flown up and hit him in the face.)

Burry gave four reasons in one go—

First, he doesn't buy the premise at all that 'today's large models are already dangerous enough to need the brakes.'

Second, once the whole industry slows down uniformly, the ones who benefit most are precisely OpenAI, Anthropic, and the others already running out front.

Then Burry abruptly pivoted to something more exciting: the IPO.

His exact words were pretty savage: an IPO needs hype and packaging, and 'we're so powerful we might become dangerous' is itself a form of hype!!

It's not that capabilities hit a wall—it's that we voluntarily hit the brakes for the safety of all humanity, ha.

If Burry only fired a shot from a distance, the next guy actually started following the money.

American investigative blogger Kevin Bass posted a long thread that was truly reasoned, sourced, and full of drama.

He straight-up turned into a su! per! sleuth!

Kevin Bass opened with a haymaker—I audited Anthropic's finances, ha~

Then he dug down through public investments, charitable donations, and institutional funding, and finally unearthed the four most dramatic points.

First bombshell: the people investing in Anthropic are also pouring money into researching 'how dangerous AI is.'

The person he focused on was Facebook co-founder Dustin Moskovitz.

Moskovitz invested in Anthropic early on, and later long-term funded AI safety organizations through his philanthropic system.

Bass cited Moskovitz's own recent public statements that they funded organizations including METR and Redwood Research.

This planted Bass's core question: the people betting on Anthropic are also paying the institutions that study how dangerous Anthropic is.

Second bombshell: METR, the third-party referee Dario recommended, overlaps with this funding network.

Dario recently called for frontier AI to undergo independent third-party evaluation, and METR was one of the institutions he highlighted.

But after sorting through it, Bass found that part of METR's funding sources and partner institutions overlap with Moskovitz's philanthropic network.

METR also works closely with Redwood, which likewise receives funding from this AI safety philanthropic ecosystem.

So Bass's harshest question really comes down to this: is the independent referee Anthropic recommends actually independent enough?

Note that METR itself says it does not accept direct funding from AI companies, but does use large amounts of free tokens provided by frontier AI companies.

Bass kept digging and found that the same AI safety philanthropic network also funded projects such as the Tarbell Center.

Tarbell trains and funds journalists to report on AI risks, and articles by its associated writers appear in outlets such as TIME, The Verge, and Science.

So Bass summed up the whole set of relationships in a highly shareable phrase—

This is what he calls the 'AI Doom Machine.'

Fourth revelation: Bass suspects this machine can keep growing on its own!!

His logic is straightforward—

The higher Anthropic's valuation, the more valuable the related philanthropic assets, the more money the AI safety ecosystem receives, and risk research and advocacy will continue to amplify, ultimately driving stronger regulation—and the ones best able to bear regulatory costs are precisely the leading AI companies.

After going around in a big circle, it comes back to giants like Anthropic.

So Bass believes there may be a self-reinforcing cycle of interests here.

On one hand calling for AI to hit the brakes, on the other rushing toward a $2 trillion IPO

What really set off this public debate was another timing point Anthropic itself can hardly avoid—the IPO.

Dario has just called on the whole industry to slow down! Yet Anthropic's capital machine is running at full speed...

Reuters reported that Anthropic is preparing what could be a record-breaking IPO, raising up to $100 billion at a valuation of about $2 trillion.

Nvidia is discussing investing up to $10 billion.

And Anthropic just completed a $65 billion funding round in May this year, when its valuation was still $965 billion.

Three months later, the story is already heading toward $2 trillion.

Of course, to Anthropic, the $2 trillion story still seems not big enough.

Just last month, it was preparing to serve investors an even more exaggerated pie—$30 trillion.

According to The Wall Street Journal, Anthropic is expected to tell investors that the TAM it faces in the future—its total addressable market—could exceed $30 trillion.

TAM does not mean Anthropic will actually earn $30 trillion in the future; it represents the annual revenue opportunity assuming it captures 100% of the relevant market.

In translation: it is already heading toward roughly one year of US GDP...

Moreover, Dario talks about pacing, but his compute bills show no sign of pacing at all.

Reuters disclosed that, to date, Anthropic has committed more than $100 billion to Amazon cloud services over the next decade.

There is also a Microsoft Azure partnership worth about $30 billion, likewise backed by large-scale Nvidia GPU compute.

In hand: $65 billion in funding received, hundred-billion-dollar compute agreements signed, and a $2 trillion IPO continuing to advance.

No truth in what it says, and it is anxious that it cannot keep up—no wonder it resents DeepSeek's open source...

As the old saying goes, no one likes saying one thing and doing another.

Has Silicon Valley become collectively desensitized to the words 'for your own good'??

Statement

(Statement: This article represents only the author's views and does not represent Sina's position.)

Source

Sina Tech