Sign up to save tools and stay up to date with the latest in AI
bg
bg
1

Google's generative AI can now analyze hours of video | TechCrunch

May 14, 2024 - techcrunch.com
Google has announced the private preview of a new version of its flagship AI model, Gemini 1.5 Pro, at the Google I/O 2024 developer conference. The updated model can analyze up to 2 million tokens, double the previous maximum, making it the largest commercially available model. This allows it to analyze larger files and maintain the context of recent conversations without veering off-topic. Developers can join a waitlist to try the new model, which has been enhanced to improve code generation, logical reasoning, planning, and understanding of audio and images.

For less demanding applications, Google is launching Gemini 1.5 Flash, a smaller and more efficient model designed for high-frequency generative AI workloads. This model is suitable for tasks like summarization, chat apps, and data extraction from long documents. Google also plans to introduce a feature called context caching for all Gemini models, allowing developers to store large amounts of information that the models can access quickly and cheaply. Other features, such as the Batch API and controlled generation, aim to provide more cost-effective ways to handle workloads and define model outputs.

Key takeaways:

  • Google has announced a new version of Gemini 1.5 Pro, its flagship AI model, which can now analyze up to 2 million tokens, double the previous maximum, making it the largest input of any commercially available model.
  • Google is also launching Gemini 1.5 Flash, a smaller and more efficient version of Gemini 1.5 Pro, designed for less demanding, high-frequency generative AI workloads.
  • All Gemini models will soon be able to use a feature called context caching, which allows developers to store large amounts of information for quick and cost-effective access.
  • Google is introducing a new feature called controlled generation, which allows users to define Gemini model outputs according to specific formats or schemas, potentially leading to further cost savings.
View Full Article

Comments (0)

Be the first to comment!