Josh Woodward, who currently leads Google Labs and oversaw the successful debut of NotebookLM, a tool that transforms text into podcast-style audio, will take over Hsiao’s role. The leadership shuffle underscores Google’s push to refine its AI offerings and capitalize on its research breakthroughs with more user-friendly applications.
A New Chapter for Gemini
Google DeepMind CEO Demis Hassabis framed the move as a necessary step to sharpen the company’s focus on the "next evolution" of the Gemini app. In a staff memo, he confirmed Woodward would continue leading Google Labs while shaping Gemini’s future.
Hsiao, in her own farewell message, described her tenure as "Chapter 1" of the Bard/Gemini story and expressed confidence in Woodward’s ability to steer "Chapter 2." After a brief hiatus, she plans to return to Google in a different capacity.
A 19-year Google veteran, Hsiao played a pivotal role in accelerating the company’s AI response following OpenAI’s ChatGPT debut, which disrupted the tech industry. Despite Google’s early contributions to transformer-based AI models—the very architecture behind ChatGPT—the company initially hesitated to release its chatbot publicly due to concerns over unpredictability.
From Early Stumbles to Breakthroughs
Google’s cautious approach didn’t shield it from early missteps. Bard’s launch was marred by "hallucinations" (fabricated responses), and its image-generation tool sparked controversy by producing historically inaccurate depictions, such as diverse Vikings and popes. Yet, the company rebounded swiftly.
Last week, Gemini 2.5 outperformed rival models from OpenAI and Anthropic in benchmark tests and surged to the top of Chatbot Arena, a platform where users rate AI responses. Meanwhile, Woodward’s Google Labs team quietly developed experimental products like NotebookLM, which converts documents into AI-narrated discussions, and Project Mariner, an autonomous browser agent capable of navigating the web and completing tasks.
The Next Phase of the AI Race
The leadership change highlights a critical transition in AI development: the battle is no longer just about superior models but about how they’re packaged into compelling products. Startups like Cursor (for developers) and Harvey (for legal professionals) have demonstrated that well-designed AI tools can carve out lucrative niches.
Last week’s events underscored this dynamic. While Google unveiled Gemini 2.5—arguably the most advanced AI model available—OpenAI’s viral image generator stole the spotlight, proving that sleek, accessible applications often resonate more than raw technical prowess.
Multimodal AI and the Road Ahead
Google DeepMind’s early bet on multimodal AI—training Gemini on text, images, and audio—could soon pay off. Though initially slower to market, this approach enables applications like Gemini Robotics, which leverages multimodal reasoning for real-world tasks. Combined with Google’s custom AI chips, which reduce operational costs, the company is well-positioned to bring innovative products to market faster.
As the AI race enters its next phase, Google’s challenge is clear: move beyond foundational research and deliver more breakthroughs like NotebookLM and Project Mariner. With Woodward at the helm, the company appears ready to embrace rapid experimentation—and finally turn its AI potential into mainstream success.