Who created the first LLM?

There's no single inventor of the first LLM — it emerged from decades of research — but the modern era began with Google, whose researchers published the transformer architecture in 2017 that all of today's large language models are built on. That paper, Attention Is All You Need, is the usual starting point people mean.

Why isn't there a single inventor of the first LLM?

Because it was an accumulation, not a single eureka moment. Language models existed well before 2017, but they were far smaller and less capable, and the whole field rests on foundational work in neural networks stretching back decades. The honest answer depends on what you mean: the key architecture came from Google researchers, while the term and much of the scaling push are associated with OpenAI.

There's no clean "first" to point to.

What did Google's transformer actually change?

The transformer made it practical to train much bigger models on much more text — that's the shift that started everything. Before it, scaling a language model up hit a wall; afterward, size became the main lever. Google itself moved fast, producing BERT in 2018, an early and influential large model that showed how powerful the new architecture was.

The transformer didn't just improve results — it changed the trajectory of the entire field.

How did OpenAI turn that into today's LLMs?

OpenAI took the transformer and scaled it up dramatically with its GPT series, showing how surprisingly capable a big enough model could be:

• GPT-1 (2018) — the first in the line, proving the approach worked.
• GPT-2 — bigger, and coherent enough that it started drawing wider attention.
• GPT-3 (2020) — the model that put "large language model" on the map for the wider industry.
• ChatGPT (late 2022) — the product built on that lineage that finally brought the technology to the general public.

So if you want names, Google gets you the architecture and OpenAI gets you the scaling and the public breakthrough — but the real answer is decades of shared research.