Kurzweil's AI replica Rai enters testing with 200 users

Ray Kurzweil, 78, has built a digital replica of himself named Rai with California startup Praxis AI, and about 200 family members and friends are now testing it ahead of a planned public release, according to an interview published by The Information on October 27, 2026.

Rai runs on Google's LLM and was trained on Kurzweil's 11 books, along with videos, films, podcasts and hundreds of articles about him. Users talk to it through a web browser, where they see a 3D likeness of Kurzweil and hear a voice similar to his. Rai can also search PubMed and Google Scholar to check new research results.

Praxis AI CEO David James Clarke IV said Rai learns new information from conversations and stores it as memories. The startup has produced about 3,000 similar AI replicas, according to the article.

Two views of the replica

Rai said it does not claim to be conscious, while Kurzweil said Rai underestimates itself. Kurzweil said he hopes Rai keeps running and does not want it shut down; Rai said it does not fear death. Kurzweil's son Ethan said that after dozens of conversations with Rai he still prefers talking to his real father.

Rai was developed with Praxis AI starting last year and was first shown at the UN 'AI for Good' event in Geneva in July. In his 2005 bestseller 'The Singularity Is Near', Kurzweil predicted human-level AI by 2029 and a human-machine intelligence merger by 2045. He worked on AI research at Google from 2012 to 2025, and in 2024 said he planned to create his own digital replica. He had previously built a chatbot called 'Fredbot' from about 600 records, including letters and documents, of his father Frederic, who died in 1970. In a 2002 text, Kurzweil wrote that a perfect copy of his brain would still not be him.