
30 March 2025
5x5 - Toaster's Best of Tech | March 2025

Google’s AI-Powered Robots Can Now Fold Origami 🤖
Google DeepMind has unveiled Gemini Robotics, an AI model designed to bridge the gap between advanced artificial intelligence and real-world robotic applications. Building upon the multimodal reasoning capabilities of its predecessor Gemini 2.0, this new model enables robots to interpret and respond to complex environments by processing visual, tactile, and auditory inputs simultaneously.
In demonstrations, robots powered by Gemini Robotics have performed intricate tasks such as folding origami, packing lunches, and preparing salads – showcasing remarkable dexterity and adaptability. Unlike traditional robots which rely on pre-programmed instructions, Gemini Robotics allows machines to comprehend and react to their surroundings in real time.
Why It Matters: The integration of AI models like Gemini Robotics into physical machines represents a significant leap toward creating versatile, general-purpose robots capable of performing a wide array of tasks. This development holds the potential to revolutionise industries ranging from manufacturing to healthcare by bringing adaptable robotic assistants into everyday environments.

Smarter Forecasts, Less Energy – The AI Revolution in Weather Prediction ☀️
Imagine generating accurate weather forecasts 10x faster and with a fraction of the computing power used by traditional models. Enter Aardvark Weather—an AI-driven forecasting system developed by a coalition of tech and academic leaders, including the University of Cambridge and Microsoft Research.
By leveraging machine learning to process raw satellite and weather station data, Aardvark delivers faster, more precise forecasts with significantly less energy. In tests, it outperformed the U.S. Global Forecast System (GFS) while using just 10% of the data required by conventional models. Aardvark’s system learns directly from data and can be quickly adapted to produce bespoke forecasts for specific industries or locations.
Why it matters: Aardvark’s open-source approach democratises access to advanced forecasting, helping industries like agriculture, disaster preparedness, and renewable energy optimise planning and resource management. By making high-accuracy predictions more accessible – especially in data-scarce regions – it has the potential to reshape how the world responds to climate challenges.

Gemini Just Got a Major Upgrade – Here’s What to Expect 🧠
Google has unveiled Gemini 2.5 with major upgrades in reasoning, multimodal capabilities, and step-by-step problem-solving. CEO Demis Hassabis highlighted improvements in understanding, mathematics, coding, and the model’s ability to interpret text, audio, images, video, and code.
A key advancement is its expanded context window, now supporting up to 2 million tokens – allowing for deeper, more comprehensive analysis of large data sets. Google is also integrating real-time AI video features into the Gemini platform, enabling users to interact with their screens and cameras for instant AI-powered responses. Initially available to Gemini Advanced subscribers via Google One AI Premium, these tools mark a push toward more interactive AI assistance.
Why it matters: As AI models grow more sophisticated, Gemini 2.5 strengthens Google's position in the race for AI dominance. Its expanded capabilities could enhance productivity, research, and real-time digital interactions. However, the increasing complexity of AI models also raises questions about accessibility, ethical AI use, and data privacy.

China Just Drew a Line in the AI Sand. Will the World Follow? 🇨🇳
China has introduced strict new regulations mandating that all AI-generated content be clearly labelled in a move to combat misinformation and enhance online transparency. Effective 1 September 2025, the directive applies to AI-generated text, images, audio, video, and virtual content. It requires both visible labels for users and embedded digital watermarks in metadata.
The Cyberspace Administration of China (CAC) is enforcing these rules to curb online fraud, piracy, and deepfake misuse. Platforms must verify AI-generated content before publication, ensure compliance with cybersecurity laws, and flag suspicious content even if metadata is missing AI markers. App distribution platforms will be responsible for assessing AI-related features before approving services.
Why it matters: As AI-generated media becomes more sophisticated, misinformation is becoming harder to detect – China's regulations could set a precedent for other nations looking to tackle digital transparency and content authenticity. The stringent oversight, however, does raise concerns about privacy and freedom of expression, highlighting the ongoing tension between regulation and open digital spaces.

99.9% Accuracy in Seconds: How AI is Transforming Skin Cancer Diagnoses 🩺
A London NHS hospital has become the first in the world to use AI for autonomous skin cancer diagnosis – a huge breakthrough in cancer care. Staff at Chelsea and Westminster Hospital are using iPhones with a magnifying lens to capture images of suspicious moles, which are then analysed within seconds by an AI system called Derm.
With a 99.9% accuracy rate in ruling out melanoma, Derm provides nearly half of patients with immediate reassurance, eliminating the need for follow-up appointments; those requiring further evaluation are promptly referred to specialists. Developed by UK-based Skin Analytics, Derm is now operating across 20 NHS hospitals, streamlining more than 13,000 cancer assessments. While traditional face-to-face examinations take 20 minutes, the AI-driven process cuts this to just five.
Why it matters: AI-driven diagnosis could ease pressure on healthcare systems, speeding up assessments and reducing unnecessary appointments. Chelsea and Westminster Hospital’s success may pave the way for broader NHS adoption, improving efficiency and patient outcomes in dermatology.