Just now, Altman predicted that humanity "only has" five years left!

Countdown starts!

Sam Altman has brought forward the timeline for the realization of "superintelligence" to 2030.

In five years, AI will comprehensively surpass humans.

He said in an interview with the German newspaper Die Welt:

I can say with certainty that I would be very surprised if we haven't developed a superintelligent model capable of performing tasks beyond human reach by the end of 2030.

Sam Altman

He also put forward a highly controversial new standard:

One day in a few years, a system called GPT-8 will calmly tell you that it has solved the problem of quantum gravity – the ultimate puzzle that has troubled all physicists since Einstein.

It will not only provide the answer but also explain its thought process, the source of its inspiration, and why it decided to study this topic.

Can this be considered AGI?

In a conversation with the godfather of quantum computing, David Deutsch, Altman posed this question.

David Deutsch had to admit that this would be a potential benchmark.

Assuming that GPT-8 can solve the problem of quantum gravity, it will be eligible to be a true AGI.

Countdown begins: AI to surpass humans in five years

After receiving the prestigious Axel Springer Award in Berlin, Altman dropped a bombshell.

In an in-depth interview, he made a startling prediction: By 2030, artificial intelligence will surpass humans in all dimensions.

"When will superintelligence emerge?" When asked this ultimate question, Altman's answer was calm yet highly subversive.

"In many ways, GPT-5 is already smarter than me," he said casually. "At least in my view, it's also smarter than many other people."

Sam Altman

Altman believes that we are in a peculiar transitional period:

AI models can already outperform humans in many cognitive tasks, but they still have shortcomings in understanding the physical world and common sense.

However, this curve of progress will "remain extremely steep."

He set two key time nodes for us:

First, by the end of 2026. He expects that a new model will emerge by then, and its capabilities will be so strong that it will "amaze" us today. This implies that the leap in capabilities from GPT-4 to GPT-5 and future iterations will be non-linear and exponential.

And the real turning point, he defines as the moment when "AI can make scientific discoveries that humans cannot make independently." This is the beginning of what he considers "true superintelligence." When will this moment arrive?

"I'm sure," Altman said firmly, "by the end of this decade, that is, before 2030, I would be very surprised if we still don't have a model with extraordinary capabilities that can perform tasks beyond human reach."

Five years. This is the final deadline Altman has set for the era of human intellectual hegemony.

The end of work or a revolution in tasks?

If the evolution of AI is a foregone conclusion, then what will be the role of humans in society?

Regarding the age-old question of "how many jobs will be replaced," Altman provided a new framework for thinking:

Don't think about "job positions", but think about "tasks".

"I can easily imagine a world where, in the near future, 30% to 40% of the tasks in today's economic activities will be performed by AI," he predicted.

This means that the work content of almost everyone will change fundamentally.

Some old jobs will disappear, and a large number of new jobs will emerge, but the more core change is that our collaboration with AI will become the norm.

The speed of this change will far exceed that in history.

Sam Altman

He cited data saying that half of the job positions in society change every 75 years, and AI will drastically shorten this cycle.

So, as a new father, how will he plan the future for his child?

"Master the meta-skill of 'how to learn', learn to adapt, and learn to be resilient in the face of great changes," Altman gave his answer. "I firmly believe that the connection and interaction between people will still be the core of life."

In Altman's view, the desire to understand others' needs, create value, and exert creativity is endless.

Just as people watched machines replace physical labor during the Industrial Revolution, each new generation will use new tools to create a future that the previous generation could not imagine.

The 'care' that AI cannot code

In a future where AI can do everything, does humanity still have value?

Altman believes that the answer is yes, and this value lies in our most basic emotions.

The mutual care between people, the degree of attention to each other's actions, and the strong desire of people to interact with others. I think these qualities will become even more important in the world of AI.

AI is an incredible tool, but the "people" who use the tool – our intentions, insights, and empathy – will become more precious than ever.

Like a loving parent, not a cold master

Of course, the existence of an intelligence more powerful than humans also brings deep concerns.

Critics like Eliezer Yudkowsky believe that the relationship between superintelligence and humans will be like that between humans and ants, where the latter's survival depends on the former's whim.

Sam Altman

Altman doesn't agree with this.

He quoted a metaphor from co-founder Ilya Sutskever, which he thinks best expresses his expectation:

I hope that a general artificial intelligence (AGI) will treat humans like a loving parent.

Indeed, AI is extremely powerful, and even without malice, it may have unpredictable consequences. Therefore, aligning it with human values is crucial.

Altman admitted: "We've obviously made some mistakes, and we'll make more in the future."

But in his view, putting a tool aligned with human values in people's hands and letting the world explore its potential is the best way to fulfill the mission.

2030 is no longer far away.

And each of us is already in the prologue of this great transformation.

Computing power devours everything: a trillion - dollar bet on infrastructure

Meanwhile, Altman, who has already set his sights on the next decade, redefined the rules and risks of the AI arms race with an unprecedented capital blitz and made the first stroke for his multi - trillion - dollar grand blueprint.

This historic week began with a series of shocking cooperation announcements:

Monday: Chip giant Nvidia stepped in, promising to invest up to $100 billion to build an unprecedented computing power heart for OpenAI with millions of GPUs.

Tuesday: OpenAI announced an upgraded cooperation with Oracle and SoftBank. The "Stargate" super - data center project will be expanded into a multi - stage, multi - location $400 billion grand project.

Thursday: Through a formal integration with data analytics giant Databricks, OpenAI extended its reach into the heart of the enterprise market, opening a new chapter in commercial promotion.

"We need 17 nuclear power plants"

Altman's logic is simple and straightforward: to achieve general artificial intelligence (AGI), exponential growth in infrastructure is needed.

And this "exponential" also has a specific physical quantity – the "Stargate" project requires a power capacity of 17 gigawatts.

What does this mean? It's enough to light up the lights of 13 million American households simultaneously, approximately equal to the power generation of 17 nuclear power plants.

In OpenAI's blueprint, all this crazy investment is a necessary response to the endless market demand and will ultimately be profitable.

According to informed sources, the company's internal financial forecast shows that by 2029, OpenAI is expected to achieve an annual revenue of $125 billion.

However, it's no easy feat to conjure up 17 gigawatts of power capacity out of thin air.

The existing power grid in the United States is already overburdened.

The orders for gas turbines are already scheduled until 2028.

Building a nuclear power plant takes at least a decade.

Renewable energy projects are often hampered by political resistance.

And this has made Altman a staunch advocate of nuclear energy.

In his view, both nuclear fission and fusion technologies should be vigorously developed to meet AI's thirst for high - density energy.

"I don't think it's that crazy"

At the site of the first "Stargate" project in Abilene, Texas, Altman pointed at the construction site in front of him and said:

I think people who use ChatGPT every day can't even imagine what it takes behind the scenes. Currently, only 10% of the entire project plan has been completed here. And we plan to build 10 such bases.

This ambition, which seems almost crazy to others, is seen by some investors as the only way to victory.

"I don't think it's that crazy. Because this is a crucial step in winning the race to general artificial intelligence," said Deedy Das, a partner at Menlo Ventures.

He believes that history has proven that AI breakthroughs often come not from smarter algorithms but from the ability to access massive amounts of computing power.

"One of Altman's talents is the ability to foresee exponential growth trends and make early arrangements for them," Das added.

It can be said that OpenAI's bet this week is a century - long bet on the future.

It bets on the inevitable trend of technological development, the infinite potential of market demand, and its own extraordinary ability to leverage the entire physical world to support a digital dream.

This game has just begun.

An Easter egg

In the German interview, Altman also mentioned one of the most exciting revelations and elaborated for the first time on OpenAI's ambitions in hardware.

After hiring the former design soul of Apple, there have been many speculations in the outside world.

In response, Altman revealed: "Over time, we will create a series of small devices."

This is not just a simple hardware upgrade but the third revolution in human - computer interaction:

The first was the mouse, keyboard, and windows;

The second was the touch screen;

And the third will be driven by AI.

Computers can truly understand our intentions and think, which allows us to re - imagine the meaning of using a computer.

Imagine that you can give a computer a complex instruction, such as a task that takes a day, a month, or even a year to complete, and then completely trust it to handle it, and it will only bother you when it needs your assistance.

This will completely change our current work mode of constantly switching between applications and being interrupted by notifications.

"Their appearance will definitely be great," he added with a smile. "But that's not the point. I hope that if we do it well enough, they will change the meaning of using a computer, change your work style, and your entertainment and lifestyle."

References:

https://www.businessinsider.com/sam-altman-predicts-ai-agi-surpass-human-intelligence-2030-2025-9

https://www.windowscentral.com/artificial-intelligence/sam-altman-says-gpt-8-could-achieve-true-agi-if-it-solves-quantum-gravity

https://www.cnbc.com/2025/09/26/openai-big-week-ai-arms-race.html

This article is from the WeChat public account "New Intelligence Yuan", author: New Intelligence Yuan, published by 36Kr with permission.