During its annual developer conference, Google unveiled its latest artificial intelligence models, which it describes as its most lightweight and efficient yet. At Google I/O on Tuesday, the company introduced Gemini 1.5 Flash, the latest iteration in the Gemini series. According to a blog post by Google, this new model is capable of rapidly summarizing conversations, providing captions for images and videos, and extracting data from extensive documents and tables.
“At a press briefing, Demis Hassabis, CEO of Google DeepMind, stated, ‘Developers expressed a desire for something faster and more cost-effective.’ This unveiling aligns with the trend among tech companies to prioritize generative AI in product development and launches. This focus is especially significant for Google, as these new tools offer consumers more sophisticated and imaginative methods to access online information compared to traditional web search.”
On Monday, OpenAI introduced a new AI model and released a desktop version of ChatGPT, accompanied by a revamped user interface. The latest model, named GPT-4o, boasts twice the speed of GPT-4 Turbo while being half the cost, as stated by the company. Meanwhile, Google has recently unveiled an enhanced Gemini 1.5 Pro model, capable of comprehending multiple extensive documents totaling 1,500 pages or summarizing 100 emails, according to a vice president overseeing Gemini.
Sissie Hsiao, Vice President at Google and General Manager for Gemini experiences, announced that Gemini 1.5 Pro will soon be equipped to process an hour of video content or codebases containing over 30,000 lines. “You can swiftly obtain answers and insights from dense documents, such as understanding the specifics of the pet policy in your rental agreement or comparing the main arguments of multiple lengthy research papers,” Hsiao explained.
OpenAI’s recent enhancement enhances both the quality and speed of ChatGPT, enabling it to accommodate 50 diverse languages. Additionally, it will be accessible through OpenAI’s application programming interface (API), enabling developers to swiftly integrate the new model into their applications. Google’s Gemini 1.5 Pro, on the other hand, supports 35 languages and features a 2 million token window, which gauges context and determines the amount of information the model can handle at once. According to company executives, the upgraded model offers enhanced capabilities in local reasoning, planning, and image comprehension.
“In the press briefing, Alphabet CEO Sundar Pichai highlighted, ‘It provides the longest context window of any foundational model to date.’ During the event, he illustrated this with an example of a parent requesting Gemini to summarize recent emails from their child’s school. Initially, Gemini 1.5 Pro will be accessible for testing in Workspace Labs. Meanwhile, Gemini 1.5 Flash will be available for testing and deployment in Vertex AI, Google’s machine learning platform designed for developers to train and deploy AI applications.”
