Google has announced a major upgrade to its artificial intelligence (AI) models that can handle up to 1 million pieces of data in a single query, enabling users to ask complex questions about the world and get answers in seconds. The new AI model, called Gemini 1.5, is part of a suite of AI models that Google launched two months ago, and is said to be five times more powerful than its closest competitor, Anthropic’s Claude 2.1.
Gemini 1.5 can process 30,000 lines of code, 11 hours of audio, or an hour of video in one prompt
According to Google, Gemini 1.5 can analyze vastly more content than before, such as 30,000 lines of code, 11 hours of audio, or an hour of video, in a single prompt. This means that users can ask the AI to perform tasks that would normally take hours or days to complete, such as extracting information from a long video, judging a rough-cut film like a critic, or querying several companies’ financial reports at once.
Google said that Gemini 1.5 is powered by its latest PaLM 2 foundation model, which uses a technique called “mixture of experts” to efficiently gather information from different sources and domains. The company also said that it has improved its ability to detect and remove questionable or harmful content, such as hate speech, bullying, and fake reviews, using its machine learning algorithm.
Google’s CEO Sundar Pichai says the new AI model is a breakthrough that will fuel the company’s businesses
In an interview with Reuters, Alphabet CEO Sundar Pichai said that the new AI model is one of multiple “breakthroughs” that will fuel the company’s myriad businesses, especially Google Search and YouTube, which rely on AI to provide relevant and useful results to users. He also said that the new AI model will help the company attract more customers to its cloud unit, which competes with Microsoft and OpenAI, among others.
Pichai said that the new AI model is making a new manner of inquiry possible, by giving users a wider view to ask questions about the world. He said that the company has discussed internally various use cases for the new AI model, such as how a movie maker could ask the AI to judge a rough-cut film like a critic, or how a researcher could ask the AI to summarize a large corpus of scientific papers. He said that the sky is the limit for the new AI model, and that the company is excited to see what users will do with it.
Google will open its new AI model to a limited number of business customers, while any developer can build with the previous version
Google said that it will open its new AI model, Gemini 1.5 Pro, to a limited number of business customers, starting from Thursday. The company said that the new AI model is typically cost-intensive, but it expects it to be profitable for the company in the long run. The company also said that any developer can build with the previous version of the AI model, Gemini 1.0, and swap in the latest generation once available.
Google also demonstrated how the new AI model works, by showing how it could extract information from a 44-minute video in about 59 seconds, or how it could respond to a multimodal prompt, in which a user asked the AI to combine text and imagery. The company said that the new AI model is versatile and covers a broad range of content types, making it a powerful tool for users to explore the world.