At the end of 2022, ChatGPT had just appeared. It still felt like a toy.
By 2025, it had cut through the entire software industry.
Only two years sat between those moments.
In 2023, you could throw a few questions at it. The answers were rough, brittle under follow-up, like a toy that happened to talk.
In 2024, it started taking real work. AI began showing up in the daily flow of work.
By 2025, it was writing code, fixing bugs, producing tests, and drafting documentation. The software industry had nowhere left to hide.
And this transformation is only just beginning.
Software Was Only the First Cut
AI's first blade fell on its own makers.
Software was closest to AI, so software was cut first.
But the blade will not stop there. It is already moving elsewhere.
The first to panic were creators. After many Chinese game companies adopted AI art tools, concept artists were among the first to feel the cuts. Voice acting, editing, and subtitles are also disappearing piece by piece.
This is not just a Chinese story. In 2023, Hollywood writers and actors went on strike, and one of their central demands was that studios should not use AI to replace them. Meanwhile, actors in Hengdian Film City have found the work drying up.
Writers have not escaped either. Web fiction platforms have already embedded AI writing tools. AI-generated novels and audiobooks are being listed in batches. Academic journals, too, are struggling with submissions written by AI.
The list will only grow. Customer service, translation, design, legal work, finance: anything that can be written down and turned into a process is being replaced.
People who can use AI well, and jobs that can be replaced by AI, are starting to separate into different layers.
Digitization Is Horizontal; AI Transformation Is Vertical
This is not the first technological revolution. The last one was called digitization.
The two look similar on the surface. Underneath, they are very different.
Digitization is horizontal. It forces you to rebuild the foundation: change processes, replace systems, install ERP, move to the cloud, remake the company from head to toe. Many things have to move at once. The cost is high, the risk is high, and the payoff is slow. Every department has to relearn its work; every process has to be rewritten. One wrong step can drag down the whole project. From the 1990s to the 2010s, simply spreading that foundation took twenty to thirty years.
AI transformation is vertical. It does not need to move the whole chessboard. It can cut into one small link first: write a block of code, fix a bug, cover one customer-service shift. The investment is low, the risk is low, and the result appears quickly: one person, one tool, one week to see whether it works. If it fails, you pull it back. There is almost no sunk cost. Digitization took twenty or thirty years to grind through the economy. AI can reshape an industry in two or three.
Why is AI moving faster than anyone expected? Because the foundation was already there. Those decades of digitization built the data, systems, and networks. AI does not have to rebuild much. It slips into existing systems and lands directly on the part that does the work.
Can traditional industries escape? No. The slowest and most expensive part of digitization was moving the real world into systems one record at a time. AI is not picky: a photo, an audio clip, a stack of handwritten forms. It can read them directly. In other words, AI can fill in that missing foundation by itself. For industries that never fully digitized, digitization and AI transformation are now arriving in one step. The thickness of the foundation only decides who goes first. It does not decide who gets spared.
AI does not tear up the foundation. It begins by replacing a door, then a window. Before you notice, AI agents are already at work in the background.
The Smoother It Feels, the More Dangerous It Is
Instinctively, "smooth" sounds like a good word. No pain. No disruption. Reassuring.
But think about replacing a window.
When you rebuild the foundation, the whole building shakes. Everyone can see it. Even when digitization eliminated jobs, it first had to install new systems. That gap gave people a little time to breathe, switch roles, and learn a new skill.
But replacing one window does not require evacuating the building. It may not even alert anyone.
What AI smooths away is exactly that buffer. No friction. No negotiation. It simply takes over the person doing the work, without making a sound.
That is the problem. Society's buffer mechanisms, retraining, policy support, psychological preparation, were all built for slow change. They assume transformation will move like digitization: slowly enough to absorb over twenty or thirty years.
This round is too fast and too smooth. Before the buffer zone is built, the people have already been replaced.
We were not unprepared because we failed to see it. We were unprepared because this time it moved too fast to leave us time.
The Iceberg Has Arrived
Ten years from now, perhaps only one-tenth of people will still be working. The direction is already clear: there will be fewer and fewer jobs that do not need AI.
Three years ago, ChatGPT was still a toy. Today, it is chewing into every industry.
The iceberg is here. Most people still cannot see what is below the surface.
No comments:
Post a Comment