AI – Why it's different this time

First of all, I am no expert in historical analysis, no futurologist, and certainly no clairvoyant. I am simply a person who likes to get lost in thought. I must disappoint anyone expecting a highly scientific analysis of the subject here – it is a thought experiment built on my axioms of logic. Such an approach carries risks, primarily those of oversimplification and false completeness – which in turn can be an indication of an overly simple formal system on my part.

Note: This article is a thought experiment based on my axioms of logic and not a highly scientific analysis of the subject.

1) The Evolution of Tools

Historically, we humans have had one advantage: our cognitive abilities. They allowed us to conceive and use tools. These tools, in turn, enabled us to ascend to higher spheres of collective human intellect – by unlocking new food sources and the associated increase in the individual's cognitive ability, as well as through the creation of collective leisure time. While the evolutionary advancement of cognitive abilities positively influences the development of tools, the unlocking of leisure time is also multi-layered. Tools can not only increase efficiency but also have an indirect impact on life expectancy and population size – through improvements in medicine, food supply, and hygiene – as well as accelerate the creation of new tools. Boiled down, tools buy us time. Since we humans have always used this time to design new tools, we find ourselves in an explosion of collective human intellect.

Language is one of those tools that makes it easier for humans to develop new tools. By sharing ideas, we can learn from each other and build upon one another instead of constantly reinventing the wheel. Books and later the internet expanded the reach of ideas – but everything is based on language. It remains the medium in which ideas are authored.

2) AI and the Rules of the Game

The idea of artificial intelligence is not new. Humanity has been fascinated by automated systems in the past. There are various approaches to developing artificial intelligence. Aside from underlying technologies, a distinction is also made regarding the scope of impact. Specialized AI has already replaced human intellect in many areas – clear rules and goals facilitated its development. For artificial general intelligence (AGI), however, clear rules and goals are missing. Our only reference for highly developed general intelligence (humans) has yet to understand the rules of its own game. One idea that has borne fruit in the recent past is using language as the playing field for AI. The goal? Predicting the next word fragment (token) based on an input. Now we have clear rules and a goal. General intelligence in terms of effect, not in function.

Do we have human intelligence now? No. Do we "only" have another powerful tool? I would also argue: no. Tools buy us time. We humans have used this time to conceive further tools. A well-known thought is that technology does not cost jobs because it creates new ones. Essentially, this means: We use our newly gained leisure time to do other things until we make those things more efficient with new tools and gain more free time again. This cycle accelerates ever further because some tools produce a positive feedback loop (language, books, internet) and thus accelerate the conception of new tools. Add to this factors like longer lifespans and population growth. But why is artificial intelligence, even if far removed from human intelligence, more than just another tool in this cycle?

3) The Paradigm Shift

Let me briefly outline why it is different this time. We have discovered language as the basis for general artificial intelligence. However, language is not just any medium; it is the medium through which we convey our ideas – the fundamental basis of every new tool. The pattern of our ideas is reflected in the pattern of our language. We have thus created another specialized AI, but within a universal medium. General artificial intelligence does not have to be human, does not have to copy our intelligence, but only has to triumph in the area that was reserved for us as general intelligence: the conception of tools.

3.1) The End of the Cycle

A system that excels in conceiving new tools ends our familiar cycle (new field – new tool – free time – new field). Unlike any technology before, we now have a tool that creates other tools. We are losing the one fundamental task with which we fill our gained free time.

Not everyone is directly involved in conceiving new tools, but indirectly, we all are. Initially, AI will feel like another powerful tool. It increases efficiency – like many technologies before it. But the cycle will break: First, no new jobs will be created. Slowly but surely, existing jobs will be lost. Initially, this will only affect certain industries – but this will increase competitive pressure in the remaining sectors. Ultimately, with few exceptions, every industry will be affected by it.

4) Implications for the Present

Perhaps we are not there yet; perhaps current systems have not yet fully mastered the complexity of language. The point of this thought experiment, however, is a different one: We do not need to create general intelligence in the sense of replicating the human brain. I question whether we currently have the necessary know-how to do so. A specialized system operating in a universal medium (e.g., language) is sufficient to take "tool creation" off our hands, thereby initiating the end of the familiar paradigm. That is why it is different this time.

4.1) What Does This Look Like in Practice?

This thought experiment illuminates the long-term paradigm shift. But what does this mean for software development today? While AI is on its way to becoming the ultimate tool, its naive use in current practice harbors risks.

How to use these systems efficiently today – but above all, safely for production tools – and why the human architect is currently still indispensable, can be read in my practical article:

Vibe Coding