Everything tagged neural-network, newest first. Tags come from the classifier reading each item's summary; 394 tags used 25 times or more have their own page.
From Alan Turing's 1936 exploration of algorithms to Claude Shannon's 1948 bit invention, foundational computer science papers have shaped today's technology landscape, influencing everything from AI to neural networks.
Dave demonstrates training a neural network on a vintage PDP144 computer, highlighting the fundamental principles of neural networks and the role of transformers in understanding sequence data through a simple task of reversing an eight-digit sequence.
Researchers have developed a chip powered by living human neurons, significantly improving energy efficiency and learning speed over traditional silicon chips, with potential implications for AI's future.
Vector search is a technique used in modern chat systems to find and use relevant data from large document sets by embedding sentences into a numerical space to identify semantically similar content.
Neurosymbolic AI combines neural networks' pattern recognition with symbolic logic's reasoning, enabling systems to both recognize and understand, thus overcoming limitations of traditional AI models that lack reasoning capabilities.
Neurosymbolic AI combines neural networks for pattern recognition with symbolic logic for reasoning, enabling systems to both recognize and understand, thus overcoming limitations of traditional AI models that excel in recognition but lack reasoning capabilities.
A neural network consists of neurons that process input vectors by assigning weights to features, applying an activation function to create nonlinearity, and learning to detect patterns that contribute to accurate predictions.
Reinforcement learning addresses the limitations of traditional neural networks by enabling AI to learn from real-world actions and outcomes, which are often non-differentiable and cannot be optimized using standard gradient descent methods.
The 2017 paper "Attention is All You Need" introduced the transformer architecture, revolutionizing AI by enabling efficient parallel processing and context retention, replacing older sequential models like RNNs and LSTMs.
This comprehensive AI research course covers foundational mathematics, PyTorch fundamentals, neural networks, and transformers, equipping learners with the essential knowledge to understand and contribute to AI research.
Artificial intelligence has evolved over 70 years from Alan Turing's foundational Turing Test to modern neural networks, with significant milestones including the development of Lisp and early chatbots like ELIZA, illustrating the progression from programming to machine learning.
Research suggests that basic arithmetic skills are rooted in the brain's memorization pathways rather than its logical reasoning circuits.
Machine learning, a subset of AI, involves algorithms learning patterns from data to make predictions, with deep learning as a more advanced subset using neural networks, and includes supervised, unsupervised, and reinforcement learning paradigms.
Sora 2, a new video generation model, offers photorealistic AI video cloning by using low rank adaptation to efficiently retrain neural networks for personalized content creation.
Elon Musk envisions Tesla as an AI company focused on developing autonomous vehicles, with the Dojo supercomputer playing a key role in training its Full Self-Driving neural networks.
Netflix has developed an automated quality control method using neural networks to detect pixel-level artifacts in videos, reducing manual workload and enhancing storytelling by allowing creative teams to focus on content rather than technical errors.
The rapid expansion of artificial intelligence is significantly increasing energy consumption in data centers, prompting major tech companies to seek sustainable solutions for managing the power demands of large AI models.
The paper discusses using AI, specifically a self-supervised learning neural network called dreams, to interpret the vast unexplored universe of natural molecules from tandem mass spectra, potentially leading to breakthroughs in various fields like medicine and materials science.
Generative AI is evolving from large language models (LLMs) predicting tokens to language concept models (LCMs) reasoning within sentence spaces, utilizing advanced embeddings for data representation and understanding semantic relationships.
Generative AI algorithms are advancing rapidly, resembling human brain functions but differing in power, volume, and message transmission, necessitating careful training to prevent errors and enhance their utility in various domains.
The Computer History Museum has released the code for the 2012 AlexNet, a pivotal breakthrough demonstrating the potential of deep learning.
Nvidia CEO Jensen Huang's GTC 2025 keynote featured numerous announcements, including a historical reference to the impactful AlexNet neural network architecture.
Diffusion models, inspired by physical diffusion processes, use neural networks to add and reverse noise in images, enabling text-to-image generation like DALL-E-3.
The Mixture of Experts podcast discusses the new $200/month o1 Pro AI model from OpenAI, with differing opinions on its value and potential adoption among AI professionals.
The video explores the process of cloning a voice using local and open-source software, emphasizing privacy and cost-effectiveness, while providing a step-by-step guide and necessary tools for creating a personalized AI voice assistant.
Ben, Ryan, and Matt Zeiler discuss AI advancements, including transformer architecture, infrastructure needs, AI regulation, and synthetic data's importance.
The podcast "Smart Talks with IBM" explores the concept of openness in AI, featuring Rebecca Finlay discussing accountable AI governance and the importance of transparency for responsible technology scaling.
The 2024 Nobel Prize in Physics was awarded to John Hopfield and Jeff Hinton for foundational work on neural networks, highlighting the intersection of physics and computer science in advancing artificial intelligence.
AI models, inspired by the brain's neural networks, process inputs through layers of nodes to create intelligence.
Large language models store facts, like "Michael Jordan plays basketball," in multi-layer perceptrons, though understanding this storage remains complex.
Google and Tel Aviv University introduced Game and Gen, a neural network-based game engine simulating real-time graphics without coding.
Mixture of Experts is a machine learning approach that activates specific subnetworks for tasks, improving efficiency in large models.
Back propagation is a key algorithm in neural networks that adjusts weights and biases to minimize error, enhancing the network's learning and adaptability through error correction and optimization techniques like gradient descent.
The discussion explores the potential for detecting AI-generated text, particularly from ChatGPT, by subtly altering outputs to identify them as AI-created, rather than relying on inefficient neural network training, while maintaining text readability and quality.
