A profile of Character.AI, founded by two former Google researchers to offer “plausible conversation” with famous people, like William Shakespeare and Elon Musk
Cade Metz / New York Times : Tweets: @patrickerville , @nytscience , and @hickspeggy Tweets: Patrick Somerville / @patrickerville : Do me please, I have questions. https://twitter.com/... @nytscience : Experts are still debating whether the strengths of chatbot technologies will outweigh their flaws and potential for harm, but they agree on one point: The believability of make-believe conversation will continue to improve https://www.nytimes.com/... Peggy Hicks / @hickspeggy : Just wondering on the comparative degree of difficulty between those two examples? 🤔 https://twitter.com/...
Context & Ripple Effects
Two months after telling the Washington Post they had left Google's LaMDA project to open large language models up to public experimentation, Noam Shazeer and Daniel De Freitas are getting the full New York Times profile treatment — the moment a startup crosses from demo to story. The Times frames Character.AI around its signature trick: plausible conversation with famous people, from William Shakespeare to Elon Musk.
The profile lands in a fast-moving arc: by March the company would raise $200M-$250M led by a16z at a ~$1B valuation, and Platformer's later coverage of ChatGPT's human voice and Meta's AI characters would name this exact space — synthetic companions — as a candidate for a new consumer internet era. The Times' own framing captures the tension: experts agree believability will keep improving even as they debate whether strengths outweigh harm.
First-order effects
- Google loses two of the researchers behind LaMDA to a consumer product that ships what Big Tech was still keeping in the lab — character-based chatbots open to anyone, immediately.
- Users gain direct access to make-believe conversation with simulated public figures like Shakespeare and Musk, putting believability — and the harm questions attached to it — in consumers' hands rather than researchers'.
Second-order effects
- Big Tech's own character play follows: Meta's AI characters and ChatGPT's conversational polish, per later coverage, push the same companion dynamic into mainstream platforms, forcing Character.AI to compete against distribution it cannot match.
- Consumer chatbots split into two markets — utility assistants versus personality-driven companions — a split visible when industry insiders later gravitate toward Claude for feeling less robotic while Replika users build emotional bonds with their bots.
Third-order effects
- If believability keeps improving as the experts in the piece agree it will, simulated personas of real people become a standing legal and platform-governance question: who consents to being a character, and who is liable when the impersonation misleads.
- The pattern points toward a consumer internet where attention flows to synthetic personalities rather than human creators — the structural shift Platformer names, with Character.AI as its early, venture-backed data point.
The trend: Conversational AI is splitting off a consumer segment built on synthetic personalities and companions, moving faster than the labs that invented the underlying models are willing to ship.