At Google I/O 2025, Google unveiled a series of groundbreaking AI advancements. Leading the announcements was Gemini 2.5 Pro with its new Deep Think mode, which achieved top scores across mathematics, coding, and multimodal reasoning benchmarks.
In the creative domain, Google introduced Flow, an AI filmmaking tool powered by Veo 3, Imagen 4, and Gemini. Flow enables users to generate high-quality videos complete with synchronised audio, camera movements, and dialogue from simple text prompts, streamlining the video production process. This new feature is specifically aimed at influencing Hollywood and it will likely have a major impact on the industry.
One of the most exciting features is project Astra, it is an experimental AI assistant developed by DeepMind, aiming to create a universal digital helper that understands and interacts with the world around you. Unlike traditional assistants, Astra processes multimodal inputs – text, speech, and visuals – in real time, allowing it to assist proactively. Watch this demo that demonstrates its uses.
Additionally, Google launched Agent Mode for Gemini, allowing users to delegate complex, multi-step tasks to the AI. This feature integrates with Project Mariner, supporting up to 10 simultaneous tasks and a “Teach and Repeat” function, enabling the AI to learn and replicate user-demonstrated tasks. There is also now a google deep search similar to OpenAI’s deep research.
They unveiled so many new features and advancements that they published a dedicated blog post highlighting the top 100 announcements.
OpenAI has teamed up with former Apple design chief Jony Ive to create a new AI hardware device. The project, still under wraps, aims to move beyond traditional screens, offering a more natural way to interact with AI. Ive, known for designing the iPhone and iMac, brings his minimalist design approach to this venture. OpenAI CEO Sam Altman has hinted that the device could be “the coolest piece of technology the world will have ever seen”.
The collaboration is backed by significant funding, including support from the Emerson Collective, and is expected to raise up to $1 billion by the end of the year. While details are scarce, the device is anticipated to integrate advanced AI functions, such as summarising information, recognising objects, and managing complex tasks. This partnership marks a significant step for OpenAI into the consumer hardware market.
Microsoft has unveiled Aurora, a new AI model designed to transform global weather forecasting. Trained on over a million hours of atmospheric and climate data, Aurora can generate 10-day weather forecasts and 5-day air pollution predictions in under a minute, outperforming traditional models in both speed and accuracy.
Aurora’s architecture – a 1.3 billion-parameter model – enables it to capture intricate atmospheric dynamics. Operating at a high spatial resolution of 0.1° (approximately 11km at the equator), it provides detailed insights into weather patterns and extreme events.
In tests, Aurora has demonstrated superior performance in predicting extreme weather events, such as accurately forecasting the path of Typhoon Doksuri four days in advance, outperforming seven leading forecasting centres.
In the multicultural hub that Luxembourg is, it is imperative that one of the most important speeches of the year is to be understood by all. Therefore, 360CrossMedia uploaded the State of the Nation address to youtube in four different languages – English, French, German, and Portuguese. In a LinkedIn post the company’s owner, Jerome Bloch, addressed the president of the Chamber: “Our point is: ‘The technology is here and it is affordable. Dear Mr Claude Wiseler, please use it to make our democracy stronger’.”
Here is the link to the English version.
OpenAI and UAE-based G42 are developing a monumental AI data centre in Abu Dhabi, named Stargate UAE, which will span 10 square miles – larger than Monaco – and consume 5 gigawatts of power, equivalent to five nuclear reactors.
Scheduled to begin operations in 2026, the first phase will deploy 200 megawatts of capacity, utilising Nvidia’s advanced Grace Blackwell GB300 servers. This facility aims to serve as a regional hub for AI infrastructure, supporting applications across the Middle East, Africa.
The project is part of a broader initiative between the U.S. and UAE to enhance AI capabilities, with OpenAI as a primary anchor tenant. It underscores the UAE’s strategic push to become a global leader in AI and digital infrastructure.
