On May 14, 2024, the logo of the Google I/O Developer Conference can be seen at the venue in Mountain View, California.
Image Alliance | Image Alliance | Getty Images
Google is using its annual developer conference to showcase what the company calls its lightest and most efficient artificial intelligence model.
At the Google I/O conference on Tuesday, the company announced the Gemini 1.5 Flash, the latest addition to the Gemini model series.
“We’re hearing from developers that they want something faster and more cost-effective,” Google DeepMind CEO Demis Hassabis said in a news release.
The launch comes as tech companies increasingly refocus their product development and rollout around generative artificial intelligence, which is particularly important for Google as the new tools offer consumers more advanced capabilities than traditional web searches. , more creative ways to access online information.
OpenAI launched a new artificial intelligence model and a desktop version of ChatGPT on Monday, as well as a new user interface. The company says the new model, GPT-4o, is twice as fast as GPT-4 Turbo but costs half as much.
Google also announced an improvement Gemini 1.5 Professional Edition According to the vice president responsible for Gemini, the model is capable of understanding multiple large documents (1,500 pages in total) or summarizing 100 emails.
Gemini 1.5 Pro will soon be able to handle an hour of video content or a code base of more than 30,000 lines, Hsiao said.
“You can quickly get answers and insights on dense documents, such as figuring out the details of a pet policy in a rental agreement or comparing key arguments in multiple long-form research papers,” Xiao said.
OpenAI’s latest upgrade, announced this week, improves the quality and speed of ChatGPT in 50 different languages. Executives said it will also be available through OpenAI’s application programming interface (API), allowing developers to start building applications using the new model immediately.
Google says Gemini 1.5 Pro is available in 35 languages and has 2 million token windows that measure context and indicate how much information the model can process at one time. Company executives say the new model improves local reasoning, planning and image understanding.
“It provides the longest contextual window of any underlying model to date,” Alphabet CEO Sundar Pichai said at a news conference. At the event, he gave the example of a parent asking Gemini to summarize recent information from their child’s school. All emails.
Gemini 1.5 Pro will initially be tested in Workspace Labs. Gemini 1.5 Flash will be tested on Vertex AI, Google’s machine learning platform that allows developers to train and deploy AI applications.
watch: Artificial intelligence dominates annual list of disruptors
About The Author
