IBL News | New York
Yesterday, at the Google I/O developer conference, Google unveiled LearnLM, a new family of generative AI models fine-tuned “to make teaching and learning experiences more active, personal and engaging.”
LearnLM is already powering features across Google products, including YouTube, Google’s Gemini apps, Google Search and Google Classroom.
Google also plans to partner with organizations, including Columbia Teachers College, Arizona State University, NYU Tisch, and Khan Academy to see how LearnLM can be extended beyond its own products.
In addition, new to YouTube are AI-generated quizzes to academic videos. This new conversational AI tool allows users to figuratively raise their hands when watching longer educational videos, such as lectures or seminars.
Users can ask the AI to summarize the video or tell them why it’s important. If they want to test their knowledge, they can ask the AI to quiz on the topic. The AI will them give then a multiple-choice question.
These new features are rolling out to select Android users in the U.S.
Google announced it’s adding an open-source framework called Firebase Genkit, that enables developers to build AI-powered applications in JavaScript/TypeScript with Go for content generation, summarization, text translation, and images creation.
Google Keynote (Google I/O ‘24)
More relevant announcements at the Google I/O developer conference:
• Google will be adding a new 27-billion-parameter model to the Gemma 2 model, launching in June.
• Gmail is getting an AI-powered upgrade. From a sidebar in Gmail, users will be able to search, summarize, analyze attachments, like PDFs, and draft their emails using its Gemini AI technology. They will also be able to take action on emails for more complex tasks, like helping users organize receipts or process an e-commerce return by searching their inbox, finding the receipt, and filling out an online form.
• Gemini can now analyze longer documents, codebases, videos, and audio recordings than before. The new Gemini 1.5 Pro can take in up to 2 million tokens, that is, double the next-largest model, Anthropic’s Claude 3, tops out at 1 million tokens.
Two million tokens are equivalent to around 1.4 million words, two hours of video, or 22 hours of audio.
• Google is also building Gemini Nano, the smallest of its AI models, directly into the Chrome desktop client, starting with Chrome 126. This, the company says, will enable developers to use the on-device model to power their own AI features.
• Google announced Imagen 3, the latest in the tech giant’s Image generative AI model family.
• Targeting OpenAI’s Sora, Google unveiled Veo, an experimental AI model that can create minute-long 1080p video clips after a text prompt. For now, Veo remains behind a waitlist on Google Labs, the company’s portal for experimental tech, for the foreseeable future.
.
