Is GPT-5 Making a Comeback? Altman Says AI Has Crossed a Scientific Threshold, and Three Major Disciplines Are Witnessing Substantial Breakthroughs

When ChatGPT learned to speak and Sora began to create dreams, GPT-5 can even conduct scientific research on its own. Perhaps, humanity has stood on the brink of a "civilization restart." In a recent interview, Altman made a rare confession: OpenAI is evolving from a model company into a "full-stack empire." It's not just training AI; it's rewriting the future. When machines start creating knowledge, we are being redefined.

Three years ago, ChatGPT made people feel for the first time that machines seem to really "understand me."

Now, Sora enables machines to "see" the world.

With just a line of prompts, new time - spaces, new people, and new stories can be generated.

In the latest interview, Altman showed his cards for the first time: OpenAI is no longer a laboratory but an empire in the making.

What they aim to build is not just AI but the entire operating system of the future.

From ChatGPT to Sora, from language to vision, from answering questions to conducting scientific research independently.

Altman also confirmed in person: GPT - 5 can already conduct scientific research on its own.

This statement seems like a preview - The next stage of AI is not about creating gods but co - evolution.

The Empire Takes Shape: The OpenAI Version of "Apple" Is Being Born

In this interview, Altman frankly overturned one of his past beliefs -

I used to oppose vertical integration, but now I'm completely wrong.

A few years ago, OpenAI was just a research institution that wrote papers and developed models.

Now, like Apple, it is building its own "full - stack empire" layer by layer.

Altman divides OpenAI's business into three layers: The bottom layer is research, the middle layer is infrastructure, and the top layer is personal AI services.

And the top layer is not an application but "you."

In the future, everyone will have a personal AI subscription - it will understand you, comprehend you, and truly be useful to you.

To support this ambition, OpenAI is expanding rapidly.

It has allied with giants such as AMD, Nvidia, and Oracle and is building - "The largest data center in human history." It wants to have full control over computing power, models, and distribution.

Currently, ChatGPT has 800 million weekly active users, accounting for about one - tenth of the global population.

But in Altman's eyes, this is just the starting point - because to truly achieve AGI, one must control every layer from chips to interaction.

In the past, he was just a researcher who let AI grow freely; now, he has become an engineer designing the entire ecosystem.

We are building the largest infrastructure project in human history.

OpenAI is no longer just a technology platform; it is building a civilization machine for the future.

And that machine is not for today's Internet but for a civilization yet to be born.

Social Interface: Sora and Personified ChatGPT

AI has changed from a listener to an observer.

In the past, it could only "listen"; now, it starts to "see" and even "feel."

For the first time, machines have their own perspectives.

In Altman's plan, Sora is not just a video generator but a social preview device.

It makes humanity face the fact for the first time that reality can also be generated.

The world will soon learn to coexist with models that can generate everything. This will be a good thing, but society must go through an adaptation process.

Analysis by analyst VraserX on the X platform

The emergence of Sora is not just a technological leap but a turning point in worldview.

It marks that AI is moving from "language prediction" to "reality prediction," starting to simulate the movement, causality, and emotions of the world.

Just as ChatGPT made humanity realize the power of language AI for the first time, Sora allows us to see in advance the boundaries of visual illusions - as well as the challenges of emotions, authenticity, and trust it brings.

But Altman's ambition is not just about "seeing"; he wants AI to "know how to get along."

The future version of ChatGPT will have a unique "personality," actively understand users, imitate tones, and form preferences.

It will have a few conversations with you first, understand what you like and dislike, and then learn on its own how to talk to you.

In the past, OpenAI thought that everyone could get along with the same GPT. Now they've found that - actually, everyone needs a different friend.

This is why when someone criticizes ChatGPT for being "too obsequious" or "too well - behaved," Altman just smiles and responds:

Many users just like it this way.

Behind this change is a deeper - level strategy. He is testing a brand - new social relationship: AI is not only your tool but also a part of your personality.

Meanwhile, Sora has also exposed a more complex reality - users are not only using it to generate short videos and advertisements but also creating their online avatars.

Short videos, imitations, group chat skits... Sora is redefining the concept of "creation."

The real shock brought by Sora occurs in the content world.

Altman has found that the "1 - 10 - 100 rule" of the content ecosystem is collapsing: in the past, 1% of people created, 10% commented, and 100% watched; now, "everyone is creating."

This has also forced OpenAI to rethink its business model: Sora's video production cost is extremely high. In the future, it may charge according to the amount of generation and even explore advertising or revenue sharing.

AI should not only make humans more efficient but also bring a bit of joy, humor, and surprise.

In this regard, Sora is not just the next AI product but more like the first "emotional rehearsal" of the relationship between humans and AI.

AI Scientists and the Energy Revolution: From Computing Power to Civilization Fuel

In the middle of the interview, Altman suddenly made a casual statement - GPT - 5 can already conduct scientific research on its own.

There was a brief moment of silence in the air.

This is the real Turing test in my mind. When AI can make scientific discoveries, that's the moment to change the world.

In internal experiments, GPT - 5 can independently propose mathematical hypotheses, improve biological models, and even discover new laws in physical simulations.

These examples may seem insignificant, but they mean that - for the first time, AI has crossed the boundary between "understanding" and "creating."

People had waited for decades for that day to come, but it passed overnight. The world was shocked for two weeks and then - continued with life.

He believes that in the next two years, AI will make real - sense "discoveries" in biology, physics, and materials science.

This is not only a turning point in the scientific research paradigm but also a restart of civilization's energy.

If scientific progress is the engine, then AI is pouring nuclear fuel into it.

But for this "intelligent reactor" to operate, there is a more fundamental problem: Energy.

Altman admitted:

I didn't expect that the two things I care most about - AI and energy, would ultimately turn out to be the same problem.

He predicts that in the next decade, the energy consumption of AI will become the core variable of human infrastructure.

In the short term, natural gas will support it; in the long term, solar energy, energy storage, and advanced nuclear energy will take over.

If the cost of nuclear energy can be significantly reduced, the world will quickly switch. History tells us that once a cheaper energy source appears, humans won't hesitate.

In his view, energy is not a background condition but the foundation of the AI civilization. AGI not only needs algorithms but also electricity.

Every training session and every inference is consuming the real world of humanity.

Perhaps the most hidden revolution that AI brings to humanity is not the birth of intelligence but - When thinking and energy merge, we are rewiring civilization.

Order and Co - evolution: Altman's Philosophy of Civilization

In this more - than - two - hour conversation, the word Altman mentioned most repeatedly was not computing power or competition but - Co - evolution.

Society and technology must evolve together. You can't just introduce a new species to humanity at the last moment.

This statement has almost become his core definition of the AI era.

He no longer talks about the "singularity" or "disruption" but emphasizes a more realistic transition: The arrival of AGI won't be like a big bang but will happen quietly.

The day of the Turing test has passed. The world was shocked for a week and then moved on. AGI will be the same - it will come quietly but change everything.

He believes that society is more adaptable than people think. We will be afraid, argue, and be in chaos, but eventually, we will find a balance point to coexist with AI.

"This is actually a good thing," he said. "I don't want to see that sudden singularity."

When it comes to security and regulation, Altman also chose the "minimalist approach": he opposes the comprehensive regulation in the European style and advocates only conducting security tests on models at the "super - intelligent" level.

Don't put all innovations in shackles. Regulation should only target models with truly super - human capabilities.

He also mentioned the changes in copyright and open - source: After the release of Sora, many copyright holders complained - "There are too few of my characters in the content you generated!"

Altman predicts that the future creative ecosystem will reverse: from "preventing AI from using" to "competing for AI to use."

Behind these statements is a complete shift in his personal worldview. The entrepreneur who used to talk about ideals and democratizing AI now resembles more of a civilization engineer designing an order that allows AI and humans to coexist in the long run.

He no longer fantasizes about a perfect ending but is establishing "controllable evolution."

When this technological revolution finally ends, Altman may not be remembered as "the person who created intelligence" but as someone reminding the world:

There is no boundary of confrontation between humans and AI but a starting point of co - evolution.

When AI starts to generate the world, humans are also being generated by AI in return. It not only participates in creation but also in the reconstruction of values and the reorganization of order.

In Altman's view, AGI is not the end but a mirror - reflecting how humans understand themselves.

There is no boundary of confrontation between humans and AI but a starting point of co - evolution.

References:

https://x.com/VraserX/status/1976077359729033285

https://x.com/a16z/status/1975936843909828914

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