The AI Era, the LLM Era, and the Problem of Naming Our Own Times

Imagine you’re talking to a language model and you want to say something simple: that you mean the time from when all of this started. The time when a conversation with something non-human suddenly became possible in a way that had previously been reserved only for other humans. And it turns out you don’t have a single word for it. So you say: “since large language models became widespread,” or “since ChatGPT went mainstream,” or “since the turning point of 2022.” Each of these is correct. None of them is convenient. That is exactly the problem worth examining more closely.

How Humanity Has Always Named Its Eras

For millennia, humanity has organized its own history through the lens of the tools that dominated a given time. The Stone Age, the Bronze Age, the Iron Age — these are not metaphors or poetic shortcuts. They are precise technological labels that say: during this time, this tool or material defined how people lived, worked, and fought.

The pattern repeats in modern eras: the Industrial Revolution, the Age of Steam, the Electrical Age, the Atomic Age, the Information Age. Each time the schema is the same — some technology changes enough aspects of life that historians (or simply people talking about their own times) feel the need to name the period.

There is, however, an important detail in this pattern that is easy to overlook: these names were given with a delay. The term “Stone Age” did not arise when people were using stone tools — it arose when that era was already thousands of years behind them. “The Industrial Revolution” as a concept became widespread decades after the first factories began transforming the English landscape. People living through those moments had no label for their own times. We — for the first time in history — are trying to give one to ourselves, in real time.

Where the Threshold Lies

If we’re looking for a point from which to count a new era, one date appears most often: November 2022. That is when OpenAI made ChatGPT available to the general public. Within five days the product had gathered one million users. Within two months — one hundred million. No technology in history had reached that scale so quickly.

But why this particular moment, and not an earlier one? Large language models existed before 2022. GPT-3 had been available since 2020, though mainly to developers. BERT, LaMDA, PaLM — all of them preceded ChatGPT. Even earlier, systems like IBM Watson had generated similar excitement in their time. If AI was developing gradually, why do we single out one specific moment?

The answer lies not in architecture or in the model’s parameter count, but in something else entirely: democratization through natural language. ChatGPT was the first model to reach mass scale while being simultaneously accessible to anyone — without programming knowledge, without an API, without specialized expertise. All it took was typing a sentence in plain human language and receiving a reply in plain human language. That was a qualitative shift, not a quantitative one.

It is worth noting that at roughly the same time — in 2022 — Midjourney, DALL-E 2, and Stable Diffusion brought an analogous breakthrough in the domain of images. Millions of people were able for the first time to describe in words what they wanted to see, and to see it. If we are looking for a threshold, it encompasses more than just text — it encompasses the entire moment when AI became accessible through natural language as an interface.

A Survey of Candidate Names

Since the need for a name is real, it is worth examining what is currently in use.

The AI Era is the broadest and most commonly encountered label in mainstream media. It is immediately intelligible and works well as a headline. Its weakness, however, is fundamental: AI as a concept has existed since the 1950s. Expert systems, first-wave neural networks, recommendation algorithms — these were all AI too. To say “the AI era” is really to say “the era of a specific kind of AI,” without specifying which kind.

The LLM Era (Large Language Models) is technically accurate when speaking about text-based language models. For people working with technology, the term carries the right meaning. It has two drawbacks, however: it is opaque to a broader audience, and more importantly, it excludes generative visual models, which were equally important to the cultural breakthrough of this period.

The Generative AI Era solves that problem — it covers both text and image, and is more precise than “the AI era.” It appears with increasing frequency in articles and industry reports. Its weakness is length and a certain dryness — it reads more as a technical term than a cultural one.

The Conversational AI Era emphasizes what is truly central to this breakthrough from the user’s perspective: the ability to have a conversation. Not to generate, not to process — but to dialogue. This label captures the experience rather than the architecture. Its weakness is that it is not yet well established in common usage and may require explanation.

Post-ChatGPT functions differently from the others — instead of describing the technology, it points to a specific historical moment as a reference point. It works similarly to “post-war” or “after the Industrial Revolution” — it does not define what something is, but from when we begin counting. It is legible both to humans and to language models, which understand this context without additional explanation. Its limitation: ChatGPT is a product name belonging to one specific company, which may sound anachronistic over time — much as if the electrical age were called “the Edison era.”

The Problem of Naming Your Own Era

Here we touch on something that goes beyond terminology. For the first time in history, humanity is trying to name its own era before that era has ended — and possibly before it has properly begun. We are in the middle of it.

This creates fundamental difficulties. We do not yet know which feature of this moment will turn out to be truly defining. Will it be the scale of models? Accessibility through natural language? The impact on the labor market? A change in the way people think and write? We are asking for a name without yet knowing the answer to the question: what will this time actually turn out to be?

Historians of the future will have it easier. They will see from a distance which thread of this period proved most important and will weave a name from it. We, meanwhile, are operating under conditions of incomplete information, trying to describe something we are immersed in.

And perhaps that is the most important observation: the very need to name this time — which you feel when you’re talking to a model and searching for a short phrase for “from when all of this started” — is proof that something has genuinely changed. Eras are not created by decree. They arise when enough people begin to feel that something is different from what it was before.

A Question Without an Answer

None of the proposed names is perfect. The AI Era is too broad. The LLM Era is too technical. The Generative AI Era is precise but cold. Post-ChatGPT is concrete but based on a product name. The Conversational AI Era is apt but poorly established.

Perhaps the right name has not yet been coined. Perhaps it will emerge not from the technology side, but from the experience side — from what people will feel when they look back on this time. Or perhaps it already exists in some language, in some article that has not yet had the chance to become canonical.

What do you call it? And what — in your view — was the moment when something new began?