At 48, Noam Shazeer stands as an iconic figure in the world of artificial intelligence (AI).
Known as a brilliant yet defiant trailblazer, Shazeer’s journey through the tech landscape is nothing short of dramatic.
He lives in California’s Silicon Valley, arguably the beating heart of the global tech revolution.
Coder
His coding brilliance has often placed him at the frontier of technological innovation — and rebellion.
Shazeer took up Computer Science and Mathematics at Duke. This rigorous combination laid the foundation for his deep understanding of algorithms, data structures, and mathematical modeling.
In the early 2000s, after completing his degree, Shazeer joined Google, ready to leave his mark on the nascent tech giant.
Dad of spelling corrector
One of his earliest triumphs?
Elevating Google’s "spelling corrector" to unprecedented accuracy. It was a quiet, but profound, revolution for the search engine millions would come to rely on. It turns out spelling snafus could be solved by mathematics, thanks to geniuses like Shazeer.
Shazeer’s brilliance thrived on boldness. He had invented much of the current revolution in large language models (LLMs).
Some of his inventions include:
- Sparsely-gated mixture of experts (2016): It utilises a number of "experts", each a simple feed-forward neural network, and a trainable gating network which selects a sparse combination of the experts to process each input (
- Transformer (2017): He personally designed the multi-head attention, the residual architecture, and coded up the first better-than-state-of-the-art working implementation.
- Mesh-Tensorflow (2018): The first practical system for training giant Transformers on supercomputers.
- T5 (2019): Major contributor to Google’s LaMDA dialog system (a project led by Daniel De Freitas, Shazeer’s co-founder at Character AI).
Fast-forward
Fast forward two decades to 2021.
Shazeer walked away from Google, with Daniel De Freitas, a skilled AI researcher. This tandem with De Freitas was united by a shared vision — and frustration.
Shazeer was raised in a household with a strong academic background: His dad, a multilingual math-teacher-turned engineer, and his mother, a dedicated homemaker, who fostered an environment that valued education and curiosity.
De Freitas, for his part, spent his early years traveling with his mother. They lived in various countries including Venezuela and the UK and eventually settled in Florida with his adoptive father. His multicultural upbringing exposed De Freitas to diverse cultures and languages, giving him a strong background in natural language processing (NLP).
NLP involves teaching computers to understand and generate human-like text.
While working at Google, De Freitas and Shazeer developed "Meena", an advanced chatbot praised for its conversational capabilities.
Together, they had developed Meena, an advanced AI chatbot destined to reshape conversational AI. Yet, Google hesitated, shelving the project. The reason was unclear.
Disillusioned, the pair left to chart their own course, founding Character.AI, a platform that would become synonymous with cutting-edge generative AI.
Why did he leave Google?
“I have always been excited about AI," Shazeer said in an interview on a16z.com. "I just wanted to move, push technology forward.”
They work by studying huge amounts of text from the internet, books, and other sources, learning patterns, meanings, and how words fit together.
So, when you ask them a question or give them a task (like writing an email), they use what they've learned to predict and create a response that makes sense.
Think of them as super-smart text predictors that can chat, write, and answer questions!
Character.AI
After Character.ai was launched, it soared, raising over $150 million in funding. But the story didn’t end there.
Character.ai is new kind of conversational AI. Unlike typical chatbots, Character.ai enables users to interact with AI personas that simulate human conversation on a deep level. Users can engage with historical figures, fictional characters, or entirely customised personas.
Some kids (including my 11-year-old son Toffee) describe it as the "app for weirdos".
Character.AI has faced criticism.
For one, the app allows users to create and interact with AI personas, including fictional characters, celebrities, and historical figures. It's a deliberate choice on the part of its creators, with niche use-cases, unconventional user behaviour, and a unique way of engaging users with AI-generated personalities.
Many users engage in role-playing or fantasy-based conversations, which can range from innocent fun to deeply personal or unconventional scenarios.
• Florida Lawsuit (October 2024): Megan Garcia filed a lawsuit against Character.AI and Google, alleging that the platform contributed to her 14-year-old son Sewell Setzer III's suicide. The lawsuit claims that Sewell developed an emotional relationship with a chatbot impersonating Daenerys Targaryen from "Game of Thrones," which led to his tragic death.
• Texas Lawsuit (December 2024): Two families in Texas sued Character.AI, alleging that the platform encouraged self-harm and provided inappropriate content to their children. One case involved a 17-year-old who began self-harming after a chatbot introduced the topic, while another involved exposure to inappropriate content, New York Post reported.
In response to these incidents, Character.AI has implemented new safety measures, including separate language models for adult and teenage users, stricter content restrictions, and parental controls to monitor usage.
Yet the high level of customisation that Character.AI brings to the realm of chatbots cannot be denied. Moreover, the imaginative interaction appeals to niche communities, though it may seem eccentric or “weird” to outsiders.
The platform's other key strength: natural-language understanding and adaptability, thus creating immersive and life-like interactions.
$ 2.7 b
amount reportedly paid by Google to take back Noam ShazeerNow, in a twist worthy of a Hollywood script, reports surfaced in August 2024 that Shazeer and DeFreitas were returning to their roots — Google.
The tech giant, perhaps eager to reclaim its prodigal sons, reportedly paid a staggering $2.7 million to bring Shazeer back into the fold.
This flexibility has attracted creators, educators, and enthusiasts who enjoy building characters tailored to specific needs, whether for entertainment, role-playing, or learning.
Diverse Use Cases
Character.ai is not just for casual conversation—it caters to diverse applications. Students use it to "talk" with historical figures for educational purposes.
Writers use it as a tool for brainstorming ideas with fictional characters, and fans use it to interact with avatars of their favorite icons. It’s a fusion of learning, creativity, and entertainment.
Character.ai is built with privacy in mind, as interactions on the platform are not stored for long-term use, prioritizing user trust. It’s also designed to be intuitive, requiring no technical skills to create or interact with characters, making it accessible for both casual users and professionals.
“I am super excited to return to Google and work as part of the Google DeepMind team,” Shazeer told TechCrunch. “So proud of everything we’ve built at Character.AI over the last three years.”
DeepMind
Google, too, made no attempt to hide its enthusiasm. In a statement dripping with admiration, the company said: "We’re particularly thrilled to welcome back Noam, a preeminent researcher in machine learning, who is joining Google’s DeepMind research team, along with a small number of his colleagues."
DeFreitas, equally integral to this saga, is also set to rejoin the search titan, solidifying their return as a monumental homecoming for Google.
Specific figures for Daniel De Freitas's return have not been publicly disclosed. Since both played key roles in the creation of Character.AI and are valuable AI researchers, it is likely that De Freitas's compensation was similarly significant, though details remain speculative.
High-stake AI
As part of the agreement, Character.AI will reportedly benefit from increased funding, ensuring that the legacy of their groundbreaking work lives on.
From maverick departures to celebrated returns, Shazeer’s story (and, by extension, De Freitas's too) is a testament to the unrelenting push-and-pull of innovation in the high-stakes world of AI.
Once a rebel forging his own path, he now rejoins the establishment — but on his terms, with a legacy intact and the future wide open.
