Samsung has unveiled its new S25 series, focusing on AI integration with Google’s Gemini. The devices look similar to their predecessors but offer new AI features by combining Google’s Gemini with Galaxy AI. This AI can process complex requests across app boundaries. Technical updates include the Snapdragon 8 Elite chip, reduced weight, and larger vapor chambers for cooling. All models now have 12 GB of RAM and will receive updates for seven years. Prices remain constant: S25 starts at €899, S25+ at €1149, and S25 Ultra at €1449. A new addition, the Galaxy S25 Edge, has been announced as a particularly slim variant of the smartphone, but technical details or prices are not yet known. The S Pen in the Ultra model loses its Bluetooth functions.
Elon Musk is questioning the large AI project presented by the US President at the White House. Musk wrote on his platform X, “They don’t actually have the money.” Trump had announced that ChatGPT developer OpenAI, along with Oracle and Softbank, would invest $500 billion in new AI data centers. Musk and OpenAI CEO Sam Altman are in a growing conflict. Musk was once among the OpenAI founders but later left. He has since gone to court to stop OpenAI’s transformation into a profit-oriented company. Altman’s company accuses Musk of trying to take control of OpenAI. Altman invited Musk to visit the construction site of the first data center.
Graylark Technologies has closed public access to its geolocation AI tool Geospy. The tool analyzes information on photos, such as vegetation, building styles, and road surfaces, to determine locations. It was developed for authorities, including law enforcement. Previously, it was publicly accessible, as reported by US magazine 404 Media. In response to the report, Graylark closed public access. While practical for identifying locations in old vacation photos, it also poses significant possibilities for misuse.
Google has invested another billion dollars in AI startup Anthropic, known for its Claude models. In fall 2023, Google had already invested two billion in the company. Alongside Amazon, which contributed eight billion, Google is likely the largest investor. Anthropic expects an additional two billion from Silicon Valley venture capitalists and is currently valued at about $60 billion. This move could be strategic for Google to challenge the dominant duo of OpenAI and Microsoft in the chatbot and AI API market.
Anthropic CEO Dario Amodei announced a new generation of AI assistants, “Virtual Collaborators,” at the World Economic Forum’s annual meeting in Davos. These are expected to work more autonomously than current systems, performing complex tasks like writing, testing, and documenting code independently. Amodei anticipates significant progress this year and plans to release a new language model with improved reasoning in the coming months. He emphasizes the importance of export controls against China and sees AI development as a societal challenge with fundamental questions about economic organization and human purpose.
OpenAI is reportedly working on the o4 model, following the yet-to-be-released o3 model. Chief Product Officer Kevin Weil announced in Davos that the next iteration is already in training. The development cycle is becoming shorter: from GPT-3 to GPT-4 took 18 months, while o1 to o3 took only three months. Each new reasoning model improves the AI’s ability to understand complex relationships and reason logically. The o4 model is expected to raise the bar even higher, highlighting the rapid development pace of reasoning models.
OpenAI is set to release the “Operator” AI agent as a new ChatGPT feature for browser control this week, according to a source from The Information. The system can independently perform computer tasks like programming or booking travel, as well as offer task categories like dining and events, delivery, and shopping. Within each category, users are presented with suggested prompts. When a prompt is entered, a mini-screen opens in the chatbot, displaying a browser and the Operator agent’s actions. The agent may ask for additional information, such as time and number of people for a restaurant reservation. ChatGPT users can take control of the screen at any time while the Operator works. Tasks can also be saved and shared with other users. OpenAI CEO Sam Altman sees AI agents as the next major growth horizon for artificial intelligence.
AI scraper bots have been causing significant problems for websites like Linux Weekly News (LWN.net) since the beginning of the year. The bots create a kind of DDoS attack by overloading servers with hundreds of IP addresses simultaneously. LWN founder Jonathan Corbet reports that only a small portion of traffic comes from real readers. The bots ignore “robots.txt” and stay individually below security system thresholds. Operators now need to invest time in defense measures to keep their sites online. Other platforms like pagure.io are also affected. The situation is described as “untenable” and requires new defense strategies, such as blocking entire subnets. Corbet humorously considers feeding the bots texts promoting LWN subscriptions as a solution to world problems.
The Hytron cleaning robot is designed to clean restrooms and toilets using artificial intelligence. The robotics startup Primech AI uses Nvidia technology like the Cuda Deep Neural Network Library for improved decision-making and TensorRT for optimizing neural networks. The latest version of the robot includes an Nvidia Jetson Orin Super SoM, suitable for robust AI and robotic applications and designed to operate energy-efficiently. The robot is said to autonomously clean toilets more efficiently, according to Primech AI. Hytron is intended for use by large customers, such as airports, hospitals, and shopping centers, to automate restroom cleaning. The robot uses AI to navigate tight spaces and perform various cleaning tasks autonomously. Initially, the robot will be tested in a pilot project with investment company Golden Rim Investment, which owns three hotels with about 260 rooms in Japan.