[Google I/O 2024 New Update] Google Gemini

Google announced an update on Gemini at the “Google Annual Developers Conference (I/O)” held in Mountain View, California, on the 14th. It announced the Gemini 1.5 Flash, a lightweight model for speed and efficiency, and the Gemma 2, a next-generation open model, and unveiled a new AI assistant, Project Astra.

Gemma 2, the next open model for responsible AI innovation, has been announced. Gemma 2 has a new architecture designed for breakthrough performance and efficiency and will come in new sizes.

The Gemma family is being extended to PaliGemma, the first PALI-3-inspired vision language model; and to evaluate the quality of the model response, we upgraded the Responsible Generative AI Toolkit to LLM Comparator.

The new 1.5 Flash, optimized for speed and efficiency

Last December, Google launched its first basic multi-mode model, the Gemini 1.0, in three sizes: Ultra, Pro, and Nano. Just a few months later, they launched the 1.5 Pro with improved performance and a groundbreaking long context window of 1 million tokens.

Google confirmed through user feedback that some applications require lower latency and lower service costs. This led to the introduction of the Gemini 1.5 Flash, a model lighter than the 1.5 Pro and designed to deliver large-scale services quickly and efficiently.

1.5 Flash is the latest addition to the Gemini family and is the fastest Gemini model offered by APIs. Optimized for large, frequent tasks, it provides cost-effective service and a breakthrough long context window.

Although it is a lighter model than 1.5 Pro, it has excellent multi-modal inference ability for vast amounts of information and provides impressive quality for its size.

Next generation of open models, Gemma 2

Google also shared a series of updates to Gemma, an open-model suite built with the same research and technology used to create the Gemini model.

Gemma 2, the next open model for responsible AI innovation, has been announced. Gemma 2 has a new architecture designed for breakthrough performance and efficiency and will come in new sizes.

Progress developing universal AI agents, Project Astra

As part of Google DeepMind’s mission to build responsible AI that helps humanity, they wanted to develop a general-purpose AI agent that can be helpful in everyday life. So, they are sharing the progress of building the future of AI assistants through Project Astra.

To be truly useful, agents need to understand and respond to a complex and dynamic world like people, and they need to accept and remember what they see and hear to understand the situation and take action. They also need to be proactive, educational, and personal so that users can talk naturally without delay or delay.

While remarkable progress has been made in developing AI systems that can understand multi-modal information, lowering response times to conversational levels is a challenging engineering challenge. Over the past few years, Google has been working to improve the way models perceive, infer, and communicate, so that the speed and quality of their interactions feel more natural.

Astra is a multimodal AI powered by the latest version of Google’s Generative AI Gemini. It comprehensively processes various data such as text, images, audio, and video to execute commands, and provides a function to respond to voice commands by analyzing objects seen through smartphone cameras.

In this DEMO video, you can see how Astra works in a natural way, such as finding the right item for the question, identifying the item and explaining its purpose, reviewing the written code, guiding it to its location and surroundings, and finding the lost item.

If you are more curious about the other AI announced by Google I/O 2024, you can check it out through this link.

Google Cloud Premier Partner, Cloocus

Cloocus is the Google Cloud Premier Partner, the highest level of Google Cloud Partners, and provides comprehensive cloud services based on Google Cloud. Specifically, Cloocus is rapidly acquiring Generative AI technologies that have recently been in the spotlight through a group of skilled Data & AI experts. If you need expert consultation regarding Generative AI deployment, please apply for expert consulting through the button below!