A strange question with an even stranger answer
What if you woke up tomorrow morning and had a specialist on hand for every task? A doctor to analyze your symptoms. A lawyer to review your contract. An engineer to write your code. A business consultant to strategize your business. And all of this — simultaneously, instantly, around the clock, for anyone in the world. Sounds like science fiction? Leading AI researchers say: this could be a reality in one to three years. Not for every task. Not perfectly. But surprisingly close. They call this scenario "the genius data center" – a data center powered by the combined intelligence of an entire nation of gifted individuals.How did AI get so good so quickly?
To understand what's happening right now, a simple analogy helps. Imagine you're teaching a child to read. First, they learn individual letters. Then words. Then sentences. Then books. And the more they read, the better they become — not just at reading, but at thinking, reasoning, explaining. Modern AI training works exactly like this — only at a speed and scale no human could ever achieve. Today's AI systems have metaphorically read more than a thousand people could in a thousand lifetimes combined.The result: AI models that were at the level of a student three years ago are now solving problems at a doctoral level—in medicine, law, mathematics, and programming.
And the crucial observation is: the more computing power and data you give the system, the better it gets — astonishingly reliable, almost like a law of nature. This is what researchers call "scaling."
When is the turning point?
Leading AI researchers, who work with these systems daily, openly say: they are convinced we will experience this turning point. Their assessment:- With a probability of about 50% it will happen in the next one to three years.
- With a probability of 90 to 95% it will happen in the next ten years.
For comparison: nobody knows exactly when the next earthquake will come. But geologists are very sure that it will come. The situation here is similar — only that the timeframe is measured in years, not centuries.
Why AI is Still Not Human
Here's something that surprises many: AI systems learn fundamentally differently from humans. A child learns that hot stove = pain by touching it once. It doesn't need a thousand repetitions. An AI system, on the other hand, needs huge amounts of data to understand similar relationships. Why is that? Because our brain doesn't start from scratch. Over millions of years, evolution has equipped us with basic instincts, reflexes and learning abilities. An AI system literally starts with random values - like a blank sheet of paper. That's why it has to read and process so much in order to perform similarly. But here's the fascinating thing: Once this training is complete, the systems simply know more than any individual human in many areas. A well-trained AI system knows more about medicine than most doctors. More about tax law than most tax consultants. More about history than most historians. Not because it "thinks" better - but because it has read more.The Misunderstanding About Programming
A concrete example that explains a lot: programming. For some time now, AI has been writing the majority of code in some companies. Many conclude from this: AI will soon replace all programmers. But that is a misunderstanding. There are several very different stages:- AI writes 90% of the lines of code — that sounds like a lot, but says little. A compiler also "writes" all lines of code.
- AI handles 90% of a developer's tasks — that's a much stronger statement.
- Demand for developers is falling by 90% — that's the economic consequence, which is still further away.
Two curves you should know
To understand the next few years, a simple picture helps: there isn't one development, but two. Curve 1 – What AI can do: This curve reliably and steeply rises. Models improve monthly. This is well-documented and will continue. Curve 2 - What reaches the economy: This curve follows the first one, but with a delay. An example: A new AI tool for programmers can be implemented by a startup in a week. A large corporation with 50,000 employees needs months for this — not because the tool is worse, but because it requires contracts, data privacy reviews, IT approvals, training, and internal authorizations.This is not a weakness of the technology. This is the normal process when an innovation arrives in large organizations.
Nevertheless, this second curve will proceed faster than with any previous technology. For comparison: the internet took about 20 years to fundamentally change the economy. AI will do this significantly faster.
Why AI Companies Are Making Losses—And Why That's Not a Problem
A question that troubles many: If AI is so good and so in demand – why are the leading AI companies even making losses? The answer lies in a simple problem: you have to plan for the future. Imagine opening a restaurant. You have to build and equip the kitchen a year in advance – before you know how many guests will come. If more come than expected, you'll earn a lot. If fewer come, you've invested too much. That's exactly how data centers work. AI companies have to order capacity 12 to 24 months in advance – before they know how big the demand will be. So, the enormous losses don't mean: "These companies aren't working." They mean: "These companies are betting aggressively on the future."And the numbers so far speak for them: A leading AI lab grew from 100 million euros in annual revenue to one billion — in a single year. The following year, it grew to almost ten billion. Such growth has rarely been seen in economic history.
How many providers will there be?
Another important point: Will there be a single winner in the end, or multiple winners? The answer: probably multiple. And the model is the cloud computing market. If you want to store data in the cloud today, you have three to four major providers to choose from – Amazon, Microsoft, Google, maybe one other. Not an absolute monopoly, but also not free competition with a hundred providers. Why? Because building this infrastructure is so expensive and complex that only a few companies can manage it at all. Anyone wanting to enter the market today would not only have to invest billions but also develop years of accumulated expertise from scratch.The same applies to AI: three, maybe four major providers. Each with its own strengths. And all profitable—once growth stabilizes.
What happens after the turning point?
Assuming the "Land of Geniuses" actually becomes a reality in three years. What then? Then it will become clear that the real questions are not of a technical nature. The technology will come. The difficult questions are different:- Who benefits from this? The developers of the systems, the companies that use them—or everyone?
- How quickly do new medications and medical breakthroughs reach sick people?
- What happens to countries and regions that lag far behind technologically?
An expert puts it this way: "Economic growth will come — almost on its own. What doesn't come on its own: that everyone benefits from it."
What that means for normal people
Many people are asking: What does this have to do with me? The honest answer: A lot – but not immediately. Specifically, you can expect the following in the coming years:- Medical diagnoses are becoming faster and more precise.
- Legal and tax advice will become cheaper.
- New drugs are coming onto the market faster.
- Many routine office jobs are decreasing, but new ones are being created.
Conclusion: No reason to panic — but reason to look
The key takeaways: AI will be better than any individual human in most knowledge professions in the coming years. It will take some time until this is truly noticeable in everyday life. The crucial question is who shapes it.Historians will look back on these years one day and wonder how little most people understood of it then.
Editorial Note:
This article is based on a conversation with a leading AI researcher and business leader about the current state and future of artificial intelligence. The predictions reflect the expert's personal assessment.
Deutsch 
Deutsch (Österreich) 
Deutsch (Schweiz) 
English (Australia) 
English (Canada) 
English 
English (UK) 
हिन्दी 
العربية 
Türkçe 
Русский