Google has been a key player in AI technology over the last decade, producing groundbreaking foundation models that have become the industry’s cornerstone. As Google continues to advance these technologies responsibly, the potential for transformative uses in fields as diverse as healthcare and human creativity is enormous.
Scaling Neural Networks and the Emergence of Large Models
In the process of developing AI over the past decade, we’ve discovered that scaling up neural networks can unlock surprising and innovative capabilities. However, it’s not a simple case of “bigger is better.” The creativity involved in research plays a crucial role in building effective models.
Recent advances in model architecture and training have highlighted the importance of human feedback in the loop, the power of multimodality, and efficient model building methods. These insights serve as the foundation for the next phase of AI development, aiming to provide tangible benefits in everyday life.
Introducing Google’s PaLM 2
Building on this foundation, Google is introducing PaLM 2, the next generation language model. With improved multilingual, reasoning, and coding capabilities, PaLM 2 is a state-of-the-art language model that’s faster and more efficient than its predecessors.
PaLM 2 comes in four sizes: Gecko, Otter, Bison, and Unicorn, to suit a range of use cases. The lightweight Gecko is even designed to function on mobile devices, making PaLM 2 a versatile tool that can be fine-tuned for a wide array of products, serving more people.
PaLM 2’s Impact and Applications
At the recent I/O event, Google announced over 25 new products and features powered by PaLM 2. This demonstrates PaLM 2’s ability to bring the latest AI capabilities into everyday products, benefiting consumers, developers, and enterprises worldwide.
The versatility and power of PaLM 2 show how diverse AI models can bring tangible benefits to everyone. Google is committed to creating the best foundation models, working on both the most useful and responsible AI tools today and the future of AI models.
Google’s Future in AI: The Gemini Model
Google’s Brain and DeepMind research teams have achieved many milestone moments in AI over the last decade. These two world-class teams are now merging into a single unit to accelerate progress further. Backed by Google’s computational resources, the combined team will not only enhance the products you use every day but also responsibly lead the development of the next generation of AI models.
The team is currently working on Gemini, a new model designed from the ground up to be multimodal and highly efficient at tool and API integrations. Gemini is still in training but is already showing multimodal capabilities not seen in prior models. Like PaLM 2, once fine-tuned and rigorously tested for safety, Gemini will be available in various sizes and capabilities, catering to different products, applications, and devices for everyone’s benefit.
