on june 18, noam shazeer, a landmark figure in the ai field, officially announced on x that he was leaving google to join openai. the co-creator of the transformer architecture admitted that this decision was bittersweet but expressed great anticipation for joining the openai team. openai ceo sam altman quickly responded, calling shazeer his “number one ideal partner” whom he had long hoped to collaborate with since the company’s inception, and humorously remarked, “waited ten years—worth it.”
shazeer’s academic and engineering career is a microcosm of ai development: in 2017, as a key author of “attention is all you need,” he firmly established the transformer as the cornerstone of the large‑model era; his 2016 proposal of the sparse gated mixture-of-experts (moe) architecture continues to profoundly influence model scaling paradigms; he also played a pivotal role in developing lamda and co‑founded character.ai in 2021, bringing conversational ai to the masses. in 2024, amid an intensifying ai arms race, google brought him back to deepmind through a strategic partnership, entrusting him with the co‑leadership of gemini technology. now, less than two years later, his move once again underscores the high mobility of top talent among industry giants, while also reflecting how competition at the forefront of ai has evolved from a race of technological innovation into a full‑scale battle for elite minds.
