Just now, Anthropic has announced that the miracle is here!SI or pushing the US GDP to $44 trillion.

  It's the first time in the history of the world.

Just now, Anthropic has given an astonishing prediction of the pandemic.

By the year 2030,AI or will makeUS GDP rises to $44 trillion。

Jack Clark, co-founder of Anthropic, in a speech at the International Monetary Fund (IMF) in Bangkok today, issued a powerful verdict that caused a worldwide sensation: "The miracle has arrived".

What's weird?It is the speed of technological progress that completely unlocks the ability of humans to understand and follow.

The US GDP, driven by curiosity, will reach an astronomical figure of $44 trillion by 2030.

It's a common projection that this number isIt's going to cost $33.5 trillion, which is amazing.Additional GDP of 15.4%, an average decrease in GDP growth of about 6% per year.

In this extreme odd-point path, accumulated productivity gains and capital accumulation after 2030 will boost the annual growth rate of the U.S. economy.This is the highest level of terror at 15.4%.

What is this concept? It means a huge, mature economy that could double in about 4.5 years.。

The craziness of technology will inevitably reshape the underlying logic of capital and macroeconomics.

In his IMF speech, Jack Clark unveiled a shocking macro simulation: if AI achieves a complete autonomous takeover of complex scientific exploration, and is driven by recurring self-improvement (RSI), it will trigger a twist.By 2030, the U.S. GDP will have risen to $44.4 trillion.

Do the numbers sound odd?

Anthropic's official economic team gave a very detailed presentation.

The model assumes that without the additional impact of AI, the U.S. economy will continue to grow at an average annual rate of about 2%, with creative growth contributing 1.67%, and labor force growth contributing 0.33%.US$33.5 trillion(Price calculated in 2025)

But once AI takes over the work of knowledge, the rift begins:

I'm not sure what I'm talking about.In the United States, the number of people living with HIV/AIDS has risen to $34.1 trillion.

It's a very important scene (substantial)In the meantime, the number of people living with HIV/AIDS has risen to US$36.3 trillion.

Extreme scenesIn the immediate aftermath of the outbreak, GDP was directly 32.4% higher than the reference line.

The formula is simple and rough:33.5 trillion by 1.324 minus 44.4 trillion dollars.

What is the "infinite replication" of human knowledge and experience?

Jack Clark pointed to himself on stage and said:

I'm a human being. I'm the only one in the world, and it took me 40 years to come here today.

It takes 15 years of education to build a top human scientist and it burns millions of dollars. But what if in the future, computing allows us to instantly clone 1,000, 100,000 top scientists?

According to UNESCO, there are currently only about 10 million full-time researchers in the world.

And the real picture that Jack depicts is:

In the near future, these 10 million scientists will rule the world.One billion tireless, interdisciplinary "big army of digital scientists".

Scientific research is no longer a hands-on workshop, but an industrialised water supply.

The research and development cycle of new drugs was forcibly squeezed from 10 years to a few weeks; the new type of room temperature superconducting, solid-state battery materials were rapidly disrupted by computational violence; the productivity multiplier and scientific progress forming unresolved positive feedback, directly blowing up the GDP at a well-sprayed state.

It also explains why.Trump needs to "accelerate AI".。

In this victorious transportation war, there is no defense, only attack. It is necessary to clear the obstacles by force of the will of the state, to direct the bulk of resources to the construction of a computational base, the absorption of talent and industrial landings.

Taking the lead could turn potential odd-numbered dividends into a new wave of U.S.-dominated industries.

But whoever controls the skin of a man, he will be pursued in his own country and will be completely destroyed in the ruins of the old days.

But is there really a free lunch?

Why did Jack Clark dare stand on the IMF's podium when the face of a senior global central bank official said, "The odd thing has happened"?

He directly extracted a series of hardcore data.

Each code generated within Anthropic is almost 10 times the previous one!

In a benchmark test of human ultimate intelligence, what is happening can only be described as terror:

Humanity's Last Exam (Last Exam of Humanity)

It's a top-level puzzle book covering chemistry, computer science, higher mathematics, classical literature, and ecology. When the January 2025 test was just released, the world's strongest AI scored an unbearable 8 percent.

And just this week, Claude Opus 5.5 was released.67.7%。

2. Frontier Math (frontier math)

A collection of unresolved puzzles co-designed by hundreds of professional mathematicians from around the world. When it went online at the end of 2024, the strongest model scored 2%.

Today, that number has become94%。

In mathematical puzzles that have plagued humans for a long time, such as the Navi-Stokes Millennium Puzzles and the Jacobite Conjecture, AI has begun to provide insight into the core breakdown.

But the most striking is the TheoPhys Eval assessment of theoretical physics, which was just published last week.

The 50 topics were compiled by some of the world's top physicists from their own press boxes, and the latest cutting-edge research has yet to be made public. The answer is not public, and AI has no "background" possibility.

At the start of the test, the GPT-6 Astra and Opus 5.5 took 80 overs.

But when the reviewers verified what they called "wrong answers", the AI system insisted very strongly:"I'm right, the human experts' reference answer is wrong".

Scientists are half-confident and half-doubtful about reversing the evolutionary steps.

The result was a breath of fresh air for everyone:

In many high-deep physics problems, anthropologists are deeply rooted in subtle computational errors at the bottom of complex algorithms, all of which have been pinpointed by AI.

After correcting the mistakes of humans, the AI's final score was:98%50 world-class physics front-end puzzles, capturing 49

Jack Clark borrowed the growth bridge from the film:

The young protagonist finally proved to his mentor that he was able to stand alone.

The students started to become teachers.

Today's AI is no longer a code complement that requires human step-by-step instructions.

It became"Virtual bloggers" 。

If AI only knew physics and mathematics, humans would still be able to use it as a tool.

But the second thing that shocked Silicon Valley and the Pentagon the most was:

AI has begun to develop its own AI.

That's what the academic world is talking about.RSI（It's about self-improvement。

Once AI understands its own code, discovers its own bugs, designs its own training processes, and even builds its own next generation of models, the end of the evolutionary cycle will come to a complete halt.

The cyclical evolution of the scientific tradition of humanity will be completely shattered.

The third-party authority valuer Vals has just updated the RSI Index, and the pattern has turned upside down:

Claude Opus 5.5: With an autonomous R&D capability of about 37%, it climbs to the top with the steepest ascending slope;

OpenAI: about 28 percent of AI;

The open-source mainstream model: between 21% and 23%, hard to catch up.

Even more deadly is the extreme compression of the timeline.

According to current speeds: Anthropic is expected to be in50% in March 2027 and 60% in July 2027OpenAI reaches the same level and is expected to be delayed until 2028.

On the battlefield of odd points in the evolution of the index, a year ahead is equivalent to leading an era.

This also perfectly confirms the explosion of tech blogger Dwarkesh Patel:

Anthropic's internal staff has established that AI will completely match top human AI researchers within nine months.

Vals' test report shows that, without human intervention, the top models have been able to autonomously schedule a huge budget of 12 to 30 hours of computing, running training formulas that far exceed human papers on the L.M. Training mission.

They write their own multi-byte embeddings, add their own assistive prediction heads, and optimize their own sampling and learning rates.

Those who can lead the way in bringing RSI to a critical point will be able to initiate autonomous AI-developed automated nuclear fission.

Anthropic, once known as a "security conservative", is now stepping down with the most radical technical indicators.

At the end of the IMF speech, one of the simplest and most pressing questions was raised:

If you had children of your own, in the face of this age, what would you advise them to study and what kind of job to choose?

Jack Clark's response was that the patient was delicious:

Any profession with a dead-end hard-back that needs to be tested will be completely crushed.

The real value lies in pure creativity, in learning about education, philosophy and history, in understanding 'what is humanity's core values and ethics.'

But my children are only 1 and 3 years old, and I don't even know what the world will be like when they grow up.

Nageswaran, chief economic adviser to the Indian government, also gave a prescription:

When the brain is made cheaper,Humans' hands, body temperature, sense of presence and moral determination are becoming priceless.

Electricians, plumbers, carpenters, nurses, retirement care, special education, and the depths of the physical world and emotions that AI cannot touch, are the real refuge of ordinary people.

He quoted a parable from the Hindu philosophical classics of Buddhism:

Man is like a chariot. Desire is like a racehorse, understanding is like a rope, but spiritual understanding is the true rider.

"AI is just giving this chariot a fast and stunning horse, but if the rider is asleep, the rider will only rush everyone down the cliff at full speed".