Gemini Notebook, formerly known as NotebookLM, has gained the ability to extract data directly from books in the Google Play Books library. This feature, called Expert Intelligence, is compatible with over one hundred thousand titles available in the app, allowing users to extract excerpts or access the full content through artificial intelligence.
With this addition, Google's 'academic' artificial intelligence can perform tasks such as creating summaries, organizing charts, and even generating podcasts from original material, which helps mitigate the risks of inaccuracies and hallucinations. Although this does not completely eliminate the possibility of errors, it represents a very effective way to prevent them.
As reported by The Verge, Google Labs editorial director Steve Johnson demonstrated several use cases. Among them are the creation of a recipe book and suggesting approaches for conducting difficult conversations with colleagues, using a business management book as a basis.
By enabling direct consultation through Google Play Books, Gemini Notebook simplifies the user experience, eliminating the need to manually insert links or extensive excerpts into the AI to obtain more focused answers. Experts recommend this practice as a fundamental way to reduce failures and hallucinations, significantly increasing the credibility of the AI when preparing materials such as presentations and study plans.
Google stated that it collaborated with authors to implement specific controls. However, the company only mentioned the restriction of consulting works acquired by the user themselves, excluding access to materials shared by friends.
Recently, several physical book retailers have reported an increase in large-scale acquisitions of pre-2022 publications by AI developers. This movement suggests greater confidence in written content predating the mass dissemination of artificial intelligence, aiming to avoid what is known as AI Slop.
The new feature announced by Google reinforces this trend by 'purifying' the database analyzed by the AI during response generation. Considering the model's focus on academic applications, it is crucial that the produced content is as accurate as possible. However, it is reiterated that the possibility of errors and hallucinations is not entirely eliminated, and it is essential to always verify the information and use the materials in conjunction with the original books.