Latest news with #transformer
Yahoo
27-07-2025
- Science
- Yahoo
Cannibal robot? Scientists develop a robot that can grow and heal by eating others
This robot is not the first transformer mechanism revealed to the public, but the way it transforms is certainly novel – it grows and heals by consuming other robots. Researchers from Columbia University in the United States have developed a robot, called the Truss Link, that can detect and merge with pieces of robots nearby to fill in missing parts. "True autonomy means robots must not only think for themselves but also physically sustain themselves," Philippe Martin Wyder, lead author and researcher at Columbia Engineering and the University of Washington, wrote in a statement. Related China unveils tiny spy drone that looks like a mosquito. What other small spy drones exist? Made with magnetic sticks, the Truss Link can expand or transform from a flat shape to a 3D structure to adapt to the environment. It can also add new bits from other robots or discard old parts that are not functional anymore to increase its performance. In a video posted by the team, the robot merges with a piece nearby and uses it as a walking stick to increase its speed by more than 50 per cent. Related This new artificial muscle can move just like human muscles but it's 17 times stronger 'Gives legs to AI' Researchers named the process in which the robot self-assembles bits of other robots 'robot metabolism'. It is described as a natural biological organism that can often absorb and integrate resources. Robots like the Truss Link can 'provide a digital interface to the physical world, and give legs to AI,' according to a video produced by Columbia Engineering School. Integrated with AI, they possess great potential, experts believe. "Robot metabolism provides a digital interface to the physical world and allows AI to not only advance cognitively, but physically – creating an entirely new dimension of autonomy," said Wyder. The Truss Link could, in future, be used to help develop groundbreaking technologies spanning from marine research to rescue services to extraterrestrial life. Related Stanford engineers have taken a leaf out of nature's book to build this bird robot "Ultimately, it opens up the potential for a world where AI can build physical structures or robots just as it, today, writes or rearranges the words in your email," Wyder said. Programming robots has been a challenge for engineers; however, artificial intelligence is advancing developments in robotics. 'We now have the technology [AI] to make robots really programmable in a general-purpose way and make it so that normal people can programme them, not just specific robot programming engineers," Rev Lebaredian, vice president of Omniverse and simulation technology at Nvidia, told Euronews Next in May.
Yahoo
24-07-2025
- Science
- Yahoo
Cannibal robot? Scientists develop a robot that can grow and heal by eating others
This robot is not the first transformer mechanism revealed to the public, but the way it transforms is certainly novel – it grows and heals by consuming other robots. Researchers from Columbia University in the United States have developed a robot, called the Truss Link, that can detect and merge with pieces of robots nearby to fill in missing parts. "True autonomy means robots must not only think for themselves but also physically sustain themselves," Philippe Martin Wyder, lead author and researcher at Columbia Engineering and the University of Washington, wrote in a statement. Related China unveils tiny spy drone that looks like a mosquito. What other small spy drones exist? Made with magnetic sticks, the Truss Link can expand or transform from a flat shape to a 3D structure to adapt to the environment. It can also add new bits from other robots or discard old parts that are not functional anymore to increase its performance. In a video posted by the team, the robot merges with a piece nearby and uses it as a walking stick to increase its speed by more than 50 per cent. Related This new artificial muscle can move just like human muscles but it's 17 times stronger 'Gives legs to AI' Researchers named the process in which the robot self-assembles bits of other robots 'robot metabolism'. It is described as a natural biological organism that can often absorb and integrate resources. Robots like the Truss Link can 'provide a digital interface to the physical world, and give legs to AI,' according to a video produced by Columbia Engineering School. Integrated with AI, they possess great potential, experts believe. "Robot metabolism provides a digital interface to the physical world and allows AI to not only advance cognitively, but physically – creating an entirely new dimension of autonomy," said Wyder. The Truss Link could, in future, be used to help develop groundbreaking technologies spanning from marine research to rescue services to extraterrestrial life. Related Stanford engineers have taken a leaf out of nature's book to build this bird robot "Ultimately, it opens up the potential for a world where AI can build physical structures or robots just as it, today, writes or rearranges the words in your email," Wyder said. Programming robots has been a challenge for engineers; however, artificial intelligence is advancing developments in robotics. 'We now have the technology [AI] to make robots really programmable in a general-purpose way and make it so that normal people can programme them, not just specific robot programming engineers," Rev Lebaredian, vice president of Omniverse and simulation technology at Nvidia, told Euronews Next in May.
CBS News
24-07-2025
- Climate
- CBS News
Transformer failure causes Chicago power outage in Belmont Cragin, ComEd says
A transformer failure on Chicago's Northwest Side caused several thousand residents to lose power overnight as temperatures struggled to dip below 80 degrees. The outage happened in the city's Belmont Cragin neighborhood. ComEd said a transformer "went bad" in the area of Diversey and Cicero, causing power outages in the neighborhood. Some residents were forced outside into a warm night to get a break from the heat inside their homes. Others went to their cars to have air conditioning. Still others left the area to find relief elsewhere. One neighbor said first the lights went out and then it started to heat up inside her home. ComEd crews were at the scene to get power back on for those customers and said power had been restored as of 5:30 a.m. Thursday. Chicago is under a Heat Advisory for dangerous temperatures and humidity today. Overnight, air temperatures barely got below 80 degrees in many areas, with humidity making it feel warmer. The city of Chicago and Cook County operate cooling centers that are free for all to access if they need air conditioning. Most locations open at 9 a.m.
Geeky Gadgets
11-07-2025
- Science
- Geeky Gadgets
Learn the Secrets of Building Your Own GPT-Style AI Large Language Model
What if you could demystify one of the most fantastic technologies of our time—large language models (LLMs)—and build your own from scratch? It might sound like an impossible feat, reserved for elite AI researchers or tech giants. But here's the truth: with the right roadmap, even complex systems like GPT-style models can become accessible to anyone with curiosity and determination. The rise of LLMs has reshaped industries, from content creation to healthcare, and understanding their inner workings isn't just a technical skill—it's a gateway to shaping the future. If you've ever wondered how these models predict text, understand context, or generate human-like responses, this guide will take you from zero to confident practitioner, one step at a time. In this deep dive by Marina Wyss, you'll uncover a structured, five-step approach to mastering LLMs, starting from the mathematical foundations that power them to the advanced techniques that fine-tune their performance. Along the way, you'll explore critical concepts like neural networks, transformer architecture, and alignment strategies, gaining both theoretical knowledge and practical insights. Whether you're an AI enthusiast, a developer aiming to build innovative applications, or simply curious about how these systems work, this roadmap will equip you with the tools to navigate the world of LLMs. By the end, you won't just understand how these models function—you'll see how they can be tailored to solve real-world problems and push the boundaries of what AI can achieve. 5-Step Guide to Building LLMs Step 1: Build a Strong Mathematical Foundation Mathematics forms the backbone of artificial intelligence, and a robust understanding of key mathematical concepts is essential for working with LLMs. Mastering calculus, linear algebra, and probability equips you with the tools to comprehend how these models learn, optimize, and generalize. Calculus: Develop an understanding of gradients and optimization techniques like backpropagation, which enable models to improve during training. Develop an understanding of gradients and optimization techniques like backpropagation, which enable models to improve during training. Linear Algebra: Study tensors, matrix operations, and transformations, which are fundamental to neural network computations. Study tensors, matrix operations, and transformations, which are fundamental to neural network computations. Probability: Explore concepts such as likelihood estimation and uncertainty, which underpin decision-making in AI systems. To strengthen these skills, use resources like 3Blue1Brown's 'Essence of Linear Algebra' and 'Essence of Calculus' series, or Coursera's 'Mathematics for Machine Learning' specialization. These materials provide intuitive explanations and practical examples, making complex mathematical concepts more accessible. Step 2: Understand Neural Networks Neural networks are the foundation of deep learning and serve as the building blocks for LLMs. These computational models, inspired by the human brain, are designed to identify patterns, process data, and make predictions. Learn how neurons, layers, and activation functions work together to process and transform data inputs. Understand backpropagation, the algorithm that adjusts model weights based on errors to improve learning outcomes. Explore optimization techniques such as gradient descent, which fine-tune model performance during training. For practical learning, explore resources like 3Blue1Brown's neural networks playlist, StatQuest's deep learning series, or Andrej Karpathy's tutorials on backpropagation and training. These resources bridge the gap between theoretical knowledge and hands-on application, helping you build a strong foundation in neural networks. Guide to Building Your Own Large Language Model in 2025 Watch this video on YouTube. Master Large Language Models (LLMs) with the help of our in-depth articles and helpful guides. Step 3: Dive Into Transformer Architecture Transformers are at the core of modern LLMs, transforming natural language processing (NLP) by allowing models to process entire sequences of text efficiently. Understanding this architecture is critical for building and scaling LLMs. Attention Mechanisms: Study how self-attention allows models to focus on the most relevant parts of input sequences, improving comprehension and context handling. Study how self-attention allows models to focus on the most relevant parts of input sequences, improving comprehension and context handling. Positional Encoding: Learn how transformers capture the order of words in a sequence, a crucial feature for language understanding. Learn how transformers capture the order of words in a sequence, a crucial feature for language understanding. Scalability: Discover why transformers outperform traditional recurrent neural networks (RNNs) when handling large datasets and complex tasks. Resources such as 'The Illustrated Transformer' blog and Andrej Karpathy's GPT tutorials provide accessible explanations and practical insights into transformer architecture. These materials will help you understand how transformers power LLMs and their role in pre-training large-scale models. Step 4: Master Fine-Tuning Techniques Fine-tuning is a vital step in adapting pre-trained LLMs to specific tasks or domains. This process involves training a model on a smaller, task-specific dataset to enhance its performance in targeted applications. Learn traditional fine-tuning methods, such as adjusting weights on pre-trained models to improve task-specific accuracy. Explore advanced techniques like Low-Rank Adaptation (LoRA) and Quantized LoRA (QLoRA), which reduce computational costs while maintaining high performance. Understand the importance of domain-specific data in achieving precise and reliable results for specialized applications. Books like 'Natural Language Processing with Transformers' and courses such as 'Fine-Tuning LLMs' offer in-depth guidance on these techniques. By mastering fine-tuning, you can customize models for a wide range of applications, from chatbots to domain-specific NLP tools. Step 5: Focus on Alignment Techniques Alignment ensures that LLMs generate outputs that are helpful, ethical, and safe. This step is essential for building responsible AI systems that align with human values and expectations. Reinforcement Learning with Human Feedback (RLHF) is a widely used approach for achieving alignment. Understand how RLHF combines reinforcement learning with curated human feedback to refine model behavior and outputs. Study case studies like OpenAI's InstructGPT, which demonstrate the practical application of alignment techniques in real-world scenarios. Learn about the challenges of balancing utility, safety, and fairness in AI systems, and explore strategies to address these issues. Recommended resources include StatQuest's RLHF overview, OpenAI's 'Spinning Up in Deep RL,' and the 'InstructGPT' paper. These materials provide a comprehensive understanding of alignment strategies and their importance in responsible AI development. By following this roadmap, you can build a strong foundation in LLM development. Start with mathematical principles, progress through neural networks and transformers, and master fine-tuning and alignment techniques. With dedication and curiosity, you will be well-equipped to prototype GPT-style models and contribute to advancements in AI. Staying informed and continuously learning will ensure you remain at the forefront of this rapidly evolving field. Media Credit: Marina Wyss Filed Under: AI, Top News Latest Geeky Gadgets Deals Disclosure: Some of our articles include affiliate links. If you buy something through one of these links, Geeky Gadgets may earn an affiliate commission. Learn about our Disclosure Policy.
BBC News
06-07-2025
- Automotive
- BBC News
Abnormal load under escort to close North Yorkshire roads
A 200-tonne "supergrid transformer" will be escorted through North Yorkshire later, with several rolling road closures planned to accommodate abnormal load's route will follow the A19 from Middlesbrough through Birdforth, Thormanby and Shipton by Beningbrough before turning onto Overton 80-metre specialist vehicle will transport the transformer, which will travel at a reduced speed accompanied by a police of its size, in some areas, benches, litter bins and road signs will be temporarily removed, and parking restrictions may be put in place. When the road is single carriageway or "too narrow", the convoy will travel under a rolling road closure, meaning other vehicles may be stopped at points along the supergrid transformer is one of eight National Grid transformers to be delivered to Overton and Monk Fryston electrical substations by October 2025, as part of the ongoing Yorkshire Green McGready, National Grid project director, said: "Supergrid transformers are essential to our project to upgrade and reinforce the high-voltage energy network in Yorkshire and further afield."We are working closely with other organisations to limit as much of the potential disruption as possible, and we'd like to thank local communities for their support and understanding while we undertake this vital work."Letters have been sent to local residents directly affected by the load. Listen to highlights from North Yorkshire on BBC Sounds, catch up with the latest episode of Look North.
