Gemini Core Team Makes Rare Joint Appearance: Noam Shazeer and Other Key Figures in Public Conversation

Background

Several core leaders of Google's Gemini project recently made a rare joint public appearance for a conversation. The discussion was hosted by Logan Kilpatrick, with participants including Oriol Vinyals, Noam Shazeer, and Koray Kavukcuoglu — key leaders of the Gemini team.

Breaking Down the Lineup: Backgrounds of the Gemini Core Team Members

Participant Backgrounds

The lineup for this conversation was nothing short of stellar, bringing together some of the most influential researchers in AI:

• Oriol Vinyals: A senior research scientist at DeepMind and co-lead of the Gemini project, with pioneering contributions in sequence-to-sequence models and reinforcement learning. He was one of the key drivers behind the Sequence-to-Sequence (Seq2Seq) model — an architecture that maps input sequences to output sequences through an encoder-decoder structure, laying the technical foundation for tasks like machine translation and text summarization. Additionally, his Pointer Network creatively solved combinatorial optimization problems where the output vocabulary varies with the input. In reinforcement learning, Vinyals led the AlphaStar project — DeepMind's AI system that achieved Grandmaster-level play in StarCraft II, demonstrating the enormous potential of deep reinforcement learning in complex strategy games. His role as co-lead of Gemini means the project's technical roadmap draws on deep expertise across multiple cutting-edge domains including sequence modeling, multimodal understanding, and reinforcement learning.

• Noam Shazeer: One of the core authors of the Transformer paper, who co-founded Character.AI before returning to Google, and is a key technical driver behind the Gemini project. Shazeer was one of the eight authors of the landmark 2017 paper "Attention Is All You Need," which introduced the Transformer architecture and fundamentally shifted the paradigm of natural language processing and deep learning as a whole. The Transformer's core innovation was replacing the then-dominant Recurrent Neural Networks (RNNs) and Long Short-Term Memory networks (LSTMs) entirely with self-attention mechanisms, enabling models to process all positions in a sequence in parallel — dramatically improving training efficiency and long-range dependency modeling. In 2021, Shazeer co-founded Character.AI with Daniel De Freitas, a company focused on personalized conversational AI that allowed users to interact with various AI characters. It became one of the most engaging consumer AI products in terms of user retention. However, facing the enormous compute costs required to train large models, Character.AI struggled with commercialization. In 2024, Google brought Shazeer back through an innovative deal structure — rather than acquiring the company outright, Google obtained a technology license and recruited the core founders back — in a transaction valued at approximately $2.7 billion. This approach both avoided antitrust scrutiny risks and successfully reclaimed key talent. This "acqui-hire" model became a new trend in the AI industry in 2024, with Microsoft's similar move with Inflection AI following the same playbook.

• Koray Kavukcuoglu: Vice President of Research at DeepMind, who has long led deep learning research efforts. He has deep expertise in convolutional neural networks and deep reinforcement learning, and has been a behind-the-scenes driving force for many of DeepMind's landmark projects.

• Logan Kilpatrick: Head of AI Developer Relations at Google, serving as the host for this conversation. He previously held the developer relations lead role at OpenAI before joining Google in 2024 — a personnel move that itself reflects the active talent mobility in the AI industry.

The Strategic Significance of Gemini Team Collaboration

This public joint appearance by these leaders reflects Google's strong commitment to the Gemini project and the team's cohesion. In particular, Noam Shazeer's collaboration with other Gemini leaders following his return to Google is widely seen as a critical strategic move in the large model competition.

Gemini is Google's flagship product line in the large model race, first released in December 2023. Unlike OpenAI's GPT series, Gemini was designed from the ground up with a native multimodal architecture, capable of simultaneously understanding and generating text, images, audio, video, and code. Gemini's development integrates the research capabilities of Google DeepMind — an organization that itself was formed from the 2023 merger of Google Brain and DeepMind, a consolidation aimed precisely at pooling resources to address the increasingly fierce large model competition. Gemini has since evolved into multiple versions: Ultra (maximum performance), Pro (balanced performance and efficiency), and Flash (emphasizing speed and cost-effectiveness), each targeting different use cases. Gemini not only powers Google's consumer-facing AI assistant but is also deeply integrated into Google Cloud, Search, Workspace, and other core products — serving as the technological cornerstone of Google's "AI-first" strategic transformation.

What Industry Signals Does This Joint Appearance Send?

From this brief tweet, we can read several important signals:

First, the Gemini team's core leadership appears to have strong rapport. The tweet author used "enjoyed immensely" to describe the conversation, suggesting a positive internal team atmosphere.

Second, Google is increasing the public visibility of the Gemini team by bringing core technical leaders to the forefront — a strategy similar to OpenAI's frequent public appearances by its team. In the AI industry, the public profiles of technical leaders have become an important component of corporate brand building — Sam Altman for OpenAI and Dario Amodei for Anthropic are prime examples. Google has been relatively low-key in this regard, so having the Gemini core team appear collectively marks a clear shift in communication strategy.

Third, amid the white-hot competition in large models, team stability and leadership synergy have become core competitive advantages for AI companies. The large model competition in 2024–2025 has entered deep waters, with major players including OpenAI (GPT series), Google (Gemini), Anthropic (Claude), Meta (Llama), and several Chinese companies. In this race, the battle for top AI talent has become an even more critical bottleneck than compute and data. OpenAI experienced multiple waves of key personnel departures during 2023–2024, including co-founder Ilya Sutskever's departure and the exit of several safety team members — events that significantly impacted its brand and R&D momentum. By contrast, Google — through merging DeepMind with Google Brain, bringing back key talent like Shazeer, and maintaining core team stability — is building a more sustainable competitive advantage. The leadership team's public joint appearance is, in a sense, a clear signal to the outside world: "We are united."

Takeaway

Although the full content of this conversation has not yet been publicly released, the rare joint appearance of the Gemini core team itself sends an important message — Google is adopting a more open posture to showcase the strength and cohesion of its AI team, building momentum for the next phase of the large model competition. With the return of a Transformer paper author, the consolidation of DeepMind's research capabilities, and the continued iteration of its native multimodal architecture, the Gemini team is fielding what may be Google's strongest-ever technical roster in AI. What's worth watching next is how this team will translate the accumulated expertise of these top-tier talents into product competitiveness in the next generation of Gemini.