Google has introduced a new feature called Audio Overviews in its search engine. This experimental capability automatically generates brief audio summaries based on user queries. It is built on the new Gemini models and aims to make information more accessible for those who prefer listening over reading. Currently, the feature is available through the Labs program and is activated automatically if the system considers that a short audio would be beneficial.
To the audio reviews, Google will add hyperlinks to the sources so listeners can explore the topic in more depth. Users will also be able to leave feedback on the created summaries using “Like” or “Dislike” buttons.
On another front, ambitious engineers from Silicon Valley are organizing their army to integrate artificial intelligence into military operations. The U.S. Air Force has established a reserve unit called “201,” comprising top executives from companies such as Palantir (Shyam Sankar, CTO), the team behind Mark Zuckerberg (Andrew Bosworth, CTO), and representatives from OpenAI (Kevin Weil and Bob McGrew). All recruits have been commissioned as lieutenant colonels, with a mission to implement AI into military plans and actions, as well as to facilitate the modernization of military units.
Work for these specialists will be remote, with a minimum of 120 hours per year, and no mandatory basic training. Palantir has long collaborated with the military — since 2008, they have been implementing the Maven project worth around $480 million, focused on AI integration. They also partner with the Army through collaborations with Anduril, which supplies “smart” glasses. Despite promising not to develop military technology, OpenAI continues to participate in military projects.
A new law has been enacted in New York aimed at overseeing AI development. The RAISE Act legislation is designed to prevent potential disasters caused by the use of advanced AI models from giants like OpenAI, Google, and Anthropic. The law requires safety reports to be filed, and mandates reporting on failures and incidents — violations can result in fines of up to $30 million. The regulation will apply only to large companies spending over $100 million on computational resources.
Proponents emphasize that the new law should not hinder startups and scientific research, unlike the failed SB 1047 bill in California. However, critics such as Andreessen Horowitz and the team at Anthropic argue that the requirements are too broad and could negatively impact small businesses. The legislation currently awaits the governor’s signature.
The creative team at Topaz Labs announced Astra, a web-based tool designed to upscale videos to 4K resolution. This online service allows users to enhance video quality while preserving details, making it suitable for both professional remastering and restoring vintage recordings. Users can customize the processing level, focusing on details or sharpness, and utilize a frame rate boost up to 120 fps with smooth interpolation. Astra is ideal for restoring old films or improving modern media projects. A waiting list is now open, and interested users can submit their applications.
TikTok has launched three new AI-powered video generation tools that significantly simplify creating promotional content. The first — “Image to Video” — converts static images into short clips about five seconds long. By uploading a photo and providing a thematic prompt, marketers will receive several video options that can be combined to make longer videos. The second — “Text to Video” — transforms plain text prompts into complete videos without using images or templates. The third — “Showcase Products” — adds digital avatars holding products, trying on clothes, or demonstrating mobile apps, making ads more lively and resembling user-generated content.
All generated videos are labeled as AI-created, allowing integration into platforms like Adobe Express and WPP Open. TikTok has not disclosed which specific AI models are used for these tools.