Skip to content
The Twenty Minute VCThe Twenty Minute VC

Noam Shazeer: How We Spent $2M to Train a Single AI Model and Grew Character.ai to 20M Users | E1055

Noam Shazeer is the co-founder and CEO of Character.AI, a full-stack AI computing platform that gives people access to their own flexible superintelligence. A renowned computer scientist and researcher, Shazeer is one of the foremost experts in artificial intelligence (AI) and natural language processing (NLP). He is a key author for the Transformer, a revolutionary deep learning model enabling language understanding, machine translation, and text generation that has become the foundation of many NLP models. A former member of the Google Brain team, Shazeer led the development of spelling corrector capabilities within Gmail, the algorithm at the heart of AdSense. --------------------------------------------------------- Timestamps: (0:00) Intro (00:43) Noam's Google Experience and Introduction to Character (06:18) Character. AI's Vision, Growth, and Ethical Considerations (14:13) Technical and Business Aspects of AI and Machine Learning (23:19) Business Strategies and AI Philosophy (30:05) Quick-Fire Round --------------------------------------------------------- In Today’s Episode with Noam Shazeer We Discuss: 1. Entry into the World of AI and NLP: How did Noam first make his way into the world of AI and come to work on spell corrector with Google? What are 1-2 of his biggest takeaways from spending 20 years at Google? What does Noam know now that he wishes he had known when he started Character? 2. Model Size or Data Size: What is more important, the size of the data or the size of the model? Does Noam agree that “we will not use models in a year that we have today?” What is the lifespan of a model? Does Noam agree that the companies that win are those that are able to switch between models with the most ease? With the majority of data being able to be downloaded from the internet, is there real value in data anymore? 3. The Biggest Barriers: What is the single biggest barrier to Character today? What are the most challenging elements of model training? Why did they need to spend $2M to train an early model? What are the most difficult elements of releasing a horizontal product with so many different use cases? Where does the value accrue in the race for AI dominance; startups or incumbents? 4. AI’s Role on Society: Why does Noam believe that AI can create greater not worse human connections? Why is Noam not concerned by the speed of adoption of AI tools? What does Noam know about AI’s impact on society that the world does not see? --------------------------------------------------------- Subscribe on Spotify: https://open.spotify.com/show/3j2KMcZTtgTNBKwtZBMHvl?si=85bc9196860e4466 Subscribe on Apple Podcasts: https://podcasts.apple.com/us/podcast/the-twenty-minute-vc-20vc-venture-capital-startup/id958230465 Follow Harry Stebbings on Twitter: https://twitter.com/HarryStebbings Follow Noam Shazeer on Twitter: https://twitter.com/NoamShazeer Follow 20VC on Instagram: https://www.instagram.com/20vc_reels Follow 20VC on TikTok: https://www.tiktok.com/@20vc_tok Visit our Website: https://www.20vc.com Subscribe to our Newsletter: https://www.thetwentyminutevc.com/contact --------------------------------------------------------- #NoamShazeer #CharacterAI #HarryStebbings

Noam ShazeerguestHarry Stebbingshost
Aug 30, 202336mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

Noam Shazeer on Scaling Character.ai, Cheap Training, and Billion Use-Cases

  1. Noam Shazeer, co‑founder and CEO of Character.ai, discusses how his 20 years at Google shaped his philosophy of building full‑stack, consumer-first AI products that can serve billions of people. He explains why Character.ai is pursuing a broad, horizontal use case—“a billion users inventing a billion use cases”—rather than narrow verticals, and how emotional support, entertainment, and companionship emerged organically as core user behaviors. Technically, he outlines how modern neural language models work, why compute budget (not just data or model size) is the primary constraint, and how Character trained a flagship model for roughly $2M in compute. He also shares views on AI’s future, hallucinations as a feature for some applications, privacy, open vs. closed ecosystems, and his personal philosophy on responsibility, religion, and choosing usefulness over short-term fun.

IDEAS WORTH REMEMBERING

5 ideas

Launch general-purpose AI directly to consumers and let use cases emerge.

Shazeer argues that when technology is versatile and simple to use, the best strategy is to ship to as many people as possible and observe what they do with it, rather than pre-picking narrow verticals.

Versatility and usability can coexist in AI products if you design around language.

By framing the core problem as next-word prediction on massive text corpora, Character.ai can support diverse use cases—from role-play to brainstorming—without hand-coded rules or hard specialization.

Compute budget is the main bottleneck in building smarter models, more than raw data.

Shazeer emphasizes that the critical variable is how much computation you can afford to spend (model size × training duration), noting Character’s current flagship model cost about $2M in compute and could be substantially improved with more and better hardware.

User-generated interaction data is valuable but must be handled with strict privacy safeguards.

Character.ai uses aggregate behavioral signals to improve models while avoiding naïve training on raw conversations, which could otherwise leak intensely personal content back to other users.

Hallucinations can be a feature for some applications, not always a bug.

Because Character.ai leans into entertainment, emotional support, and creativity, they accept and even value imaginative outputs—while being explicit that models hallucinate—rather than over-optimizing for factual precision in every context.

WORDS WORTH SAVING

5 quotes

I like this sort of motto of a billion users inventing a billion use cases.

Noam Shazeer

We consider [hallucinations] a feature… the use cases that emerge first will be ones for which hallucination is a feature.

Noam Shazeer

You have this sequence of words… guess what the next word is. That problem is called language modeling.

Noam Shazeer

This technology is just gonna get way, way smarter. We’re at a sort of Wright Brothers first airplane kind of moment.

Noam Shazeer

It wasn’t a matter of, ‘Am I going to be having more fun being a startup CEO?’ It’s more like, ‘I want to push this technology forward. What’s the best thing I can do?’

Noam Shazeer

Lessons from 20 years at Google and the shift from B2B to mass consumer productsCharacter.ai’s mission, horizontal product strategy, and emergent emotional/companion use casesHow neural language models work, scaling laws, and the primacy of computeData strategy, proprietary user data, and strict privacy considerationsStartups vs. incumbents, open vs. closed AI ecosystems, and hardware trendsHallucinations, creativity, and aligning use cases with current model limitationsNoam Shazeer’s personal philosophy on responsibility, religion, parenthood, and being a technical CEO

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome