Kathleen Kenealy - Creating, building, and releasing Gemma, Google's open model family
Mar 28, 2024
auto_awesome
Kathleen Kenealy, Google DeepMind Software Engineer, discusses Gemma open models. They explore the benefits and risks of sharing advanced technology, Kathleen's STEM journey, and developing Gemma with the only team member. Topics include language models, AI Chatbots, and the evolution of technology in AI development.
Kathleen Kenealy emphasizes the importance of open-source initiatives in advancing machine learning research and fostering transparency.
Kenealy reflects on her childhood love for mathematics, discovering patterns and inspiring scientific curiosity.
Deep dives
Kathleen's Academic Journey and Passion for STEM and Arts
Kathleen Canelie recounts her academic journey, highlighting her love for mathematics, discovered while in an all-girls school. She reminisces about the intersection of math and theater in her life, which eventually led her to shift her major to computer science focusing on artificial intelligence after a transformative experience with TensorFlow. With a background in machine learning techniques for text-based applications from Stanford, she delves into her work at Google Brain on large language models and the importance of open source communities.
The Fascination with Mathematics and Connecting Math to Real-World Phenomena
Kathleen reflects on her childhood love for math and the beauty of mathematical concepts found in nature. She shares insightful teaching methods from her favorite math teacher, illustrating the inherent beauty and logic of mathematics. Kathleen's early exposure to math's natural intricacy and the discovery of patterns in various forms inspired her scientific curiosity.
Development of Large Language Models and TensorFlow
The transition to using TensorFlow at a startup marked a pivotal moment for Kathleen, where she realized the transformative potential of machine learning in understanding human behavior through mathematics. Exploring the challenges of encoding word meanings into code, she highlights Google's impactful release of Word2Vec, revolutionizing language understanding. This technological advancement laid the foundation for novel machine learning applications, sparking Kathleen's journey into large language model development.
Embracing Open Source and Responsible AI Development
Kathleen emphasizes the significance of open-source initiatives in advancing machine learning research and fostering transparency. Engaging in rigorous safety and responsibility protocols, she navigates the balance between optimism for the technology's positive impact and vigilance against potential risks. Kathleen entrusts in the collaborative effort of society to navigate the ethical considerations and ensure the responsible implementation of AI technology.
Meet Kathleen Kenealy, a Software Engineer at Google DeepMind, working on the next generation of large language models (LLMs) and open models. In the episode we talk about Gemma, Google’s newest family of lightweight, state-of-the-art open models built from the same research and technology used to create the Gemini models.
Join us as we unpack the history and the importance of open technology at Google and the advantages and risks of sharing cutting edge technology with the world. We learn how that open technology started Kathleen on her own journey into AI. Kathleen shares her experience growing her team from being the only team member at first working on Gemma, growing up loving STEM, her graduate school experience working with AI Chatbots, and more on this episode of People of AI hosted by Ashley Oldacre!