The most important AI story of the decade did not begin with ChatGPT.
It began with a research paper.
In 2017, a group of Google researchers published “Attention Is All You Need.” The paper introduced the Transformer, a new AI architecture that replaced older sequence-processing methods with a simpler and more scalable idea: attention. The authors were Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz Kaiser, and Illia Polosukhin.
At the time, it looked like a major step forward for machine translation.
In hindsight, it was much more than that.
The Transformer became one of the key foundations behind modern generative AI, large language models, AI coding assistants, enterprise AI agents, and the new wave of AI infrastructure companies. WIRED later described the eight authors as the Google employees who helped invent modern AI—and noted that all eight had since left Google.
That is why this story matters.
Google created one of the most important AI breakthroughs in history. But the people behind that breakthrough eventually spread across the AI industry, founding or joining companies such as OpenAI, Cohere, Anthropic, Sakana AI, Essential AI, Inceptive, Character.AI, and NEAR.
The Transformer was not only a technical invention.
It became a talent explosion.
It became a startup map.
It became the beginning of a new AI power structure.
Why the Transformer Was So Valuable
Before the Transformer, many language models processed text sequentially. They read words step by step, which made it harder to capture long-range relationships and harder to train efficiently at massive scale.
The Transformer changed that.
Its key idea was self-attention: the ability for a model to look across a sequence and decide which words, tokens, or signals matter most to each other. Instead of treating language as a simple chain, the model could learn relationships across the entire input.
Google’s original paper described the Transformer as a model based entirely on attention, removing recurrence and convolution. It also emphasized that the architecture was more parallelizable and required less training time than earlier approaches.
That technical change created enormous business value.
It made AI models easier to scale.
It made large language models more practical.
It helped unlock systems that could summarize documents, write code, answer questions, generate text, translate languages, analyze enterprise data, and power AI agents.
In other words, the Transformer helped move AI from narrow prediction systems toward general-purpose digital intelligence.
That is why today’s AI race is not just about chatbots.
It is about who controls the next layer of computing.
The Eight Authors Behind the Transformer
The Transformer paper had eight authors. Each one contributed to a breakthrough that would later reshape the AI economy.
Here is where they went—and why their paths matter.
1. Ashish Vaswani
From Google Brain to Adept, Essential AI, and possibly NVIDIA
Ashish Vaswani is the first-listed author of “Attention Is All You Need.” He became one of the most recognized names associated with the Transformer architecture.
After Google, Vaswani co-founded Adept AI, a startup focused on building AI systems that could use software tools and perform tasks for users. He later co-founded Essential AI with Niki Parmar. Essential AI describes its mission around building open models, open tooling, and reproducible AI pipelines for enterprise use. Its official website has listed Vaswani as CEO.
There has also been recent industry reporting in June 2026 that NVIDIA hired Vaswani and several Essential AI team members to work on its Nemotron open-source model effort. Because this appears to be a developing story, the safest wording is that Vaswani is publicly associated with Essential AI, while also reportedly linked to NVIDIA’s open-model work.
Why he matters:
Vaswani represents the bridge between research and the next generation of AI platforms. His post-Google work has focused on moving AI beyond chat into software use, enterprise workflows, and model infrastructure.
2. Noam Shazeer
From Google to Character.AI, back to Google Gemini, then OpenAI
Noam Shazeer is one of the most important AI engineers of the modern era.
At Google, he was part of the Transformer paper. Later, he left Google and co-founded Character.AI, a consumer AI company built around conversational characters and personalized chatbot experiences. In 2024, Google brought Shazeer back and appointed him as a technical lead on Gemini, after striking a licensing agreement with Character.AI.
Then, in June 2026, Reuters reported that Shazeer would leave Google again to join OpenAI. Reuters also described him as a vice president of engineering at Google and a co-lead of Gemini.
Why he matters:
Shazeer’s career shows how valuable top AI talent has become. His movement from Google to Character.AI, back to Google, and then reportedly to OpenAI is not just a career path. It is a signal of how aggressively AI labs now compete for a small number of people who understand how frontier models are built.
3. Niki Parmar
From Google to Adept, Essential AI, and Anthropic
Niki Parmar was one of the eight Transformer authors and later became a major figure in the AI startup ecosystem.
After Google, she co-founded Adept AI and then Essential AI with Ashish Vaswani. Public profiles and reporting have also associated her with Anthropic, where she has worked on reliable and interpretable AI systems. Forbes India described her as a co-founder of Essential AI and Adept AI Labs, and also as technical staff at Anthropic.
Why she matters:
Parmar’s path reflects one of the biggest shifts in AI: moving from research prototypes to reliable systems that can be used in real enterprise environments. Her work sits at the intersection of model capability, product reliability, and practical AI deployment.
4. Jakob Uszkoreit
From Google Brain to Inceptive
Jakob Uszkoreit played a key role in pushing the idea of self-attention forward.
Before the Transformer, many researchers relied heavily on recurrent neural networks and LSTMs for language processing. Uszkoreit was one of the researchers who believed attention could become the core mechanism, not just an add-on. WIRED’s account of the Transformer story highlights his role in advancing self-attention as a replacement for older sequential approaches.
After Google, Uszkoreit co-founded Inceptive, a company applying AI to RNA and programmable medicine. Handelsblatt’s 2026 conference profile listed him as Inceptive’s CEO and co-founder and described the company’s goal as enabling a new generation of medicines through AI and scalable biochemistry experiments.
Why he matters:
Uszkoreit shows that the impact of the Transformer is not limited to chatbots or search. The same deep learning principles that transformed language can also influence biology, medicine, and molecular design.
5. Llion Jones
From Google to Sakana AI in Tokyo
Llion Jones was another co-author of the Transformer paper. He is also widely associated with the paper’s now-famous title, “Attention Is All You Need.” WIRED reported that Jones suggested the title, giving the paper a memorable identity that helped it stand out in AI history.
After Google, Jones co-founded Sakana AI in Tokyo, where he serves as co-founder and CTO. Sakana AI describes itself as a frontier AI research and development company working on areas such as AI scientists, multi-agent orchestration, foundation models, and AI systems for Japan.
Why he matters:
Jones represents the next stage of AI research: what comes after today’s dominant Transformer-based systems. Sakana AI’s work focuses on new model-building approaches, multi-agent systems, and AI architectures inspired by collective intelligence.
6. Aidan N. Gomez
From Google to Cohere
Aidan Gomez was an intern at Google when he became one of the authors of the Transformer paper. Afterward, he co-founded Cohere, one of the most important enterprise AI companies in the market.
Cohere focuses on AI for business use cases, especially secure enterprise deployment, retrieval, productivity, and customized models. The company’s website lists Gomez as co-founder and CEO.
Reuters reported in 2025 that Cohere had doubled annualized revenue to $100 million and had shifted strategy toward private deployments for regulated enterprises. The report also noted Cohere’s valuation of $5.5 billion after its 2024 funding round.
Why he matters:
Gomez shows how Transformer research became commercial infrastructure. Cohere is not primarily a consumer chatbot company. It is an enterprise AI company, focused on helping organizations deploy language models securely and productively.
7. Łukasz Kaiser
From Google Brain to OpenAI
Łukasz Kaiser was already an influential researcher before the Transformer paper, with deep experience in neural sequence models and machine learning infrastructure.
He later joined OpenAI, where he has worked as a research scientist. An OpenAI Forum profile describes Kaiser as a co-author of the Transformer paper and as a researcher advancing model-based reinforcement learning and contemporary reasoning models at OpenAI.
Why he matters:
Kaiser’s path connects the Transformer era with the reasoning-model era. As AI moves from text generation toward multi-step reasoning, planning, and problem solving, researchers like Kaiser are central to the next frontier.
8. Illia Polosukhin
From Google to NEAR Protocol and user-owned AI
Illia Polosukhin was another co-author of the Transformer paper. After Google, he co-founded NEAR Protocol in 2018.
NEAR’s official materials describe Polosukhin as a former Google machine learning researcher and co-author of the Transformer paper. NEAR has since positioned itself around blockchain infrastructure, user-owned AI, and agent-oriented systems.
Why he matters:
Polosukhin’s path shows a different branch of the Transformer diaspora. While many authors went deeper into AI labs and enterprise AI, Polosukhin moved toward decentralized infrastructure, digital ownership, and the trust layer for AI agents.
Where Are the Transformer Authors Now?
The Bigger Story: Google Created the Spark, the Industry Captured the Fire
The Transformer story is not simply a story about people leaving Google.
It is a story about how innovation spreads.
Google had the research environment, talent, compute, and ambition to produce one of the most important AI breakthroughs in history. But once the Transformer was published, the idea became part of the broader AI ecosystem.
OpenAI used Transformer-based ideas to build GPT models.
Cohere built enterprise AI systems.
Anthropic focused on safer and more reliable AI.
Sakana AI began exploring new approaches to frontier AI in Japan.
Inceptive took AI architecture thinking into biology and RNA medicine.
NEAR connected AI with decentralized infrastructure and agent systems.
Essential AI focused on AI tools, open models, and enterprise workflows.
In other words, the Transformer did not create one company.
It created an industry.
Why Did So Many Authors Leave Google?
There are several reasons.
First, frontier AI became too big to remain inside one company. Once the Transformer proved its value, the opportunity space exploded. AI was no longer only about search, translation, or ads. It became relevant to software, customer support, law, finance, manufacturing, education, biotech, robotics, cloud infrastructure, and consumer products.
Second, the startup opportunity became enormous. Researchers who once worked on internal projects could now raise capital, build companies, and define entire markets.
Third, AI talent became one of the most valuable resources in technology. Recent reporting around Noam Shazeer, John Jumper, OpenAI, Anthropic, and Google shows how intense the AI talent war has become. Business Insider described this as a new “celebrity era” for AI talent, where individual researchers can move markets and reshape competitive narratives.
Fourth, large companies often move carefully when a new technology threatens existing business models. Google had strong AI research, but generative AI also challenged parts of the traditional search and advertising model. Startups had fewer legacy constraints and could move faster into new product categories.
The result was a new AI map.
Google created the architecture.
The ecosystem commercialized it.
What Is the Real Impact?
The impact of the Transformer Eight can be understood in four layers.
1. They changed the technical foundation of AI
The Transformer made attention-based architectures the center of modern AI. It helped models scale more effectively, learn richer relationships, and support a broader range of tasks.
Without the Transformer, the current pace of generative AI would likely look very different.
2. They changed the startup landscape
Many of the authors became founders or senior researchers at important AI companies. Their movement helped transfer deep model-building knowledge from one research lab into the broader market.
This is why the Transformer paper is often viewed not only as a technical document, but also as the origin story for a generation of AI companies.
3. They changed the meaning of AI talent
In the cloud era, companies competed for engineers.
In the AI era, companies compete for researchers who can design, train, scale, and align frontier models.
The Transformer authors became symbols of this shift. Their careers show that a small number of people can influence billions of dollars in enterprise value, investment strategy, and product direction.
4. They changed business strategy
For business leaders, the lesson is clear:
The winning company is not always the company that invents the technology first.
The winner is the company that can combine:
- breakthrough research
- product speed
- compute access
- data strategy
- distribution
- safety
- enterprise trust
- and world-class AI talent
That is why the Transformer story matters far beyond the AI research community.
It is a business strategy case study.
What Business Leaders Should Learn from the Transformer Eight
The first lesson is that AI advantage is mobile.
Research talent can move. Ideas can spread. Open papers can become startup ecosystems. A breakthrough inside one company can become the foundation of an entire market.
The second lesson is that productization matters as much as invention.
Google invented the Transformer. But OpenAI, Anthropic, Cohere, and others turned Transformer-based models into products, APIs, enterprise systems, and developer platforms.
The third lesson is that AI is now a full-stack competition.
It is not enough to have a model. Companies need infrastructure, data pipelines, security, governance, user experience, and workflow integration.
The fourth lesson is that the next AI winners may come from unexpected places.
A biotech company, an enterprise software company, a Japanese frontier AI lab, a decentralized infrastructure project, or an AI agent startup may all trace part of their technical DNA back to the same 2017 Google paper.
That is the power of a foundational architecture.
Final Takeaway
The Transformer paper was only a few pages long, but its consequences are still unfolding.
Eight Google researchers introduced an architecture that helped make modern generative AI possible. Then those researchers left Google and carried their expertise into new companies, new products, and new industries.
That is the real story of the Transformer.
It was not just a model architecture.
It was the beginning of an AI diaspora.
It turned researchers into founders.
It turned a paper into a platform.
And it turned artificial intelligence from a research field into one of the most important business battlegrounds of our time.
Leave a Reply