The Key Figure Behind Gemini's IMO Gold Medal Almost Became a Professional Pianist
Yi Tay, a research scientist at Google DeepMind, led the team that helped Gemini Deep Think win a gold medal at the International Mathematical Olympiad. But beyond AI, he is also an accomplished pianist who once dreamed of a career in music. This article explores his journey in AI research and his musical talent.
Article intelligence
Key points
- Yi Tay is a Google DeepMind research scientist and key contributor to Gemini Deep Think.
- He led the team that earned Gemini a gold medal at the IMO, and also contributed to physics and chemistry Olympiads.
- Yi Tay holds an Associate Diploma in Classical Piano from Trinity College London, nearly becoming a professional musician.
- After a brief stint as a startup co-founder, he returned to Google and lost 20 kg.
Why it matters
This matters because yi Tay is a Google DeepMind research scientist and key contributor to Gemini Deep Think.
Technical impact
May affect model selection, inference cost, product capability, and evaluation benchmarks.
Yi Tay, a research scientist at Google DeepMind, is widely recognized as a key figure behind Gemini's success in the International Mathematical Olympiad (IMO). Last year, he served as modeling co-captain for Gemini Deep Think, which achieved a gold-medal level performance at the IMO. The model also excelled in the physics and chemistry Olympiads earlier this year. Yet, Tay's talents extend far beyond artificial intelligence; he is also a highly skilled classical pianist.
In a recent Instagram video, Tay shared a 14-year-old recording of himself performing Chopin's "Fantasie-Impromptu" — a technically demanding piece known for its rapid arpeggios and emotional intensity. At the time, he was an undergraduate at Nanyang Technological University (NTU) in Singapore, juggling exams and piano practice. He later earned an Associate Diploma in Classical Piano Performance from Trinity College London, a credential equivalent to semi-professional level. Tay has said that in an alternate universe, he would have pursued a career as a professional musician.
Tay's professional journey in AI began at Google Brain, where he made significant contributions including UL2 — a framework unifying encoder-decoder and autoregressive pretraining paradigms — and DSI, a generative retrieval paradigm that redefines search as direct document identifier prediction. These technologies have been deployed in systems like YouTube recommendations and Spotify. He was also a modeling co-lead for PaLM-2 and contributed to Flan-2.
In 2023, Tay left Google to co-found Reka AI, a startup that rapidly developed a GPT-4-level multimodal model and briefly ranked in the top five on the LMSYS leaderboard. However, the entrepreneurial grind took a toll: he consumed five cups of coffee daily, ate two takeout meals, and gained 15 kg while his wife was pregnant. After 639 days, he decided to return to Google DeepMind, expressing genuine happiness to be back with the research infrastructure and collaborators he loved.
Since returning, Tay has shed about 20 kg, resumed badminton (even winning a mixed doubles tournament at Google Singapore), and spent quality time with his family, watching his daughter grow from one to two years old. He remains passionate about music and hopes to pick up classical piano again after retirement. While the AI world benefits from his research, perhaps the piano world awaits a late-blooming star.