Google I/O 2023 saw the arrival of several new AI-powered features and improvements that are making their way to the company’s products and services, and the firm has already begun a public preview of some of these features. Project Tailwind was one of them; it was an AI-powered note-taking targeted at improving consumer note-taking. The company says that Project Tailwind will let users take their freeform notes and automatically organize and summarize them using artificial intelligence (AI) technology.
In essence, it’s an AI notebook that has been trained on your documents and that you can use as a sort of personal instructor or writing companion. Although Google presented it as a tool for students, it may be quite useful to anyone who reads a lot of text in everyday life.
According to Google, the AI tool is aimed at college students, writers, researchers, analysts, and lawyers reviewing a case.
“Like a real notebook, your notes and your sources power Tailwind,” said Josh Woodward, a senior director of product management at Google. “How it works is you can simply pick the files from Google Drive, and it effectively creates a personalized and private AI model that has expertise in the information you give it. We’ve been developing this idea with authors like Steven Johnson and testing it at universities.”
Innovative AI-Powered Content Generation and Multilingual Capabilities
It’s not a novel idea, as James Vincent of The Verge reminds out. A recent AI addition to the note-taking app Notion allows it to generate blog posts, job descriptions, and other types of content in the manner of the user. Leaked language models are being trained on Google Drive data by the open-source community elsewhere.
However, Tailwind also acts as a demo for PaLM 2, Google’s successor for PaLM, its large language model (LLM) for understanding and creating multilingual content. PaLM 2 is used to power several of Tailwind’s features via a newly introduced API.
It’s important to note that numerous companies are currently investigating this topic. Why not fine-tune AI language models using user input when they can already be taught to handle particular kinds of data? This approach is already being taken by the note-taking tool Notion, and the open-source AI scene is full of developers refining leaky language models utilizing the data in their Google Drive.
As an example, suppose that you are a professor who has been tasked to write a study guide for university students using a particular book on astronomy. Reading the entire book will take some time in the real world. However, Project Tailwind allows you to access a soft copy from Google Drive and ask that the AI tool generate a study guide that highlights key concepts.
Google says that the tool may also generate quiz questions or question suggestions for reading comprehension quizzes. Additionally, it can cite any sources found within the book and respond to questions in natural language on the notes.