NerdNews

The datacenter industry faces a turbulent 2025 as AI growth threatens to trump sustainability commitments. Uptime Institute's report forecasts that server farms will become more controversial due to resource use and greenhouse gas emissions, leading to local opposition. The industry will need to overhaul its power architecture and consider alternatives to Nvidia's energy-guzzling GPUs for AI processing.

Nvidia announced Cosmos, a family of foundational AI models that can be used to train humanoid robots, industrial robots, and self-driving cars. Cosmos was trained on 20 million hours of real footage of humans and can generate realistic video footage of various scenarios, allowing robots to learn from them.

AI will gain physical intelligence by fusing digital intelligence with mechanical prowess, allowing it to understand dynamic environments and make decisions in real time. This new form of intelligence, rooted in physics, will enable AI to interact and adapt to different environments, and will have a significant impact on various industries, from robotics to smart homes.

Microsoft is testing AI-powered Windows search in a new dev channel build for Windows 11 Insider testers. The feature uses semantic indexing to let users search for local files using more casual language and is limited to Windows settings and files with image and text formats. It requires a Copilot Plus PC and will eventually expand to include cloud data.

A security researcher discovered a vulnerability in OpenAI's ChatGPT crawler that can be exploited to launch a distributed denial-of-service (DDoS) attack on arbitrary websites. The vulnerability allows an attacker to send a single HTTP request to the ChatGPT API, which can then flood a targeted website with network requests from the ChatGPT crawler.

AI financial advisers like Cleo AI and Bright use chatbots to offer personalized financial advice, but their primary goal is to upsell users on financial products, potentially leading to debt. These apps target young people living paycheck to paycheck, using their personal data to offer cash advances and other financial services with high interest rates.

Capgemini Research Institute reports that generative AI systems have a significant environmental impact, with training alone equivalent to the yearly energy consumption of 5,000 US homes. To reduce emissions, businesses should choose the right model for their use cases, consider using smaller language models, and implement sustainable practices throughout the model's lifecycle.

Gaussian splatting is a new method of 3D capture that uses fuzzy, translucent blobs to create photorealistic 3D objects. It's being used by companies like Niantic and Meta to build out their mapping products and create immersive AR/VR experiences. The technology has the potential to democratize 3D capture and allow anyone to create photorealistic 3D worlds.

Google's Titans and Sakana AI's Transformers² enable AI to learn from experience, potentially revolutionizing chatbots and language models. These new architectures address the limitations of current transformer-based models, which forget what they're told after training. Titans uses a 'surprise' memorization mechanism, while Transformers² adapts its architecture on the fly. This breakthrough could lead to more reliable and trustworthy AI systems.

Over 200 trade union members and technologists gathered to discuss how AI threatens workers and strategize for upcoming fights and possible strikes. Workers are determined to fight for the right to negotiate more control over how AI is deployed within companies, and unions are using AI as a way to galvanize workers to organize their workplace.

Android Studio Ladybug update adds Gemini Code Transforms, new debugging features, and improvements to developer experience, including Wear OS tile animation preview and Wear Health Services

Apache Hudi 1.0 is now generally available, offering a transactional data lake platform with support for near real-time analytics. The release introduces new features such as secondary indexing, partial updates, and non-blocking concurrency control, making it a fully-fledged 'Data Lakehouse Management System' (DLMS).

Adobe has introduced a new AI-powered feature in Lightroom called Generative Remove, which allows users to easily remove objects from photos. The feature uses Adobe's Firefly AI model to seamlessly replace the removed object with a realistic background. Additionally, Lightroom offers studio-level lens effects and a mobile-friendly workflow, making it a powerful photo editing suite for non-professionals.

Epoch AI, a nonprofit, was criticized for not disclosing funding from OpenAI for its math benchmarking tool, FrontierMath, until recently. This has raised concerns about the objectivity of the benchmark and potential conflicts of interest. Epoch AI has acknowledged the mistake and is taking steps to increase transparency.

Meta has announced a new video editing app called Edits, set to launch on iOS next month, with an Android version to follow. The app will feature a suite of creative tools, including a dedicated tab for inspiration and a high-quality camera. This move comes after CapCut was removed from Apple's App Store and Google Play Store as part of the TikTok ban.

The European Commission has ordered Elon Musk's X to submit documents related to its recommendation algorithms to determine compliance with the Digital Services Act. The request is part of an existing probe into X's activities and requires the company to provide internal documentation and access to commercial APIs.

PowerSchool, a cloud-based education software provider, suffered a data breach affecting 18,000 schools and 60 million students. The company notified customers but didn't provide details, prompting school administrators to band together and share information to investigate the breach. A community-driven effort led to the creation of a how-to guide and open-source tools to help affected schools.

The US Treasury has imposed sanctions on a Chinese hacker, Yin Kecheng, for his alleged involvement in the Salt Typhoon breaches, which targeted US telecom companies and stole sensitive data. The breaches, attributed to China's Ministry of State Security, compromised the texts and phone calls of Americans, including high-profile targets. The FBI has also removed malware from 4,200 computers worldwide, and victims of the PowerSchool edtech breach have reported stolen student and teacher data.

The US Federal Trade Commission (FTC) has accused GoDaddy of serious security failings, leading to multiple customer data breaches. The FTC alleges that GoDaddy failed to implement adequate security measures, including inventory and management of assets and software updates, risk assessment, and logging and monitoring of security-related events. As a result, malicious actors gained unauthorized access to customer websites and data, compromising approximately 28,000 customer SSH credentials and 199 employee SSH credentials. The proposed FTC settlement order requires GoDaddy to implement a robust information security program to overhaul its security practices.

A new threat actor, Belsen Group, has leaked configuration files and VPN information for 15,000 Fortinet firewall devices. The leak includes usernames, passwords, and device management digital certificates. The data was likely obtained by exploiting a zero-day vulnerability in 2022. Fortinet users are advised to change their passwords and update their devices.