Artificial Intelligence
Follow Artificial Intelligence in the Factlen app and get an alert when there is news. Get the app
Explainer
How 8-bit Integer Quantization Reduces LLM Memory Footprint by 75% with Minimal Accuracy Loss
6 sources · 45d ago
Explainer
How FlashAttention Bypasses the GPU Memory Bottleneck to Enable Long-Context AI
5 sources · 45d ago
Explainer
How a Generator and a Discriminator Compete to Create Realistic AI Output
8 sources · 46d ago
More in Artificial Intelligence
ExplainerMachine Learning
How Generative AI Maps the Joint Probability Distribution of Data
Deep Learning Practitioners 50%Classical Statisticians 30%AI Safety Researchers 20%
ExplainerActivation Steering
How Activation Steering Modifies AI Behavior Without Retraining
Interpretability Researchers 35%AI Safety Auditors 35%Model Deployers 30%
Research BriefLabor Reinstatement
The Economic Finding of Job Reinstatement: How AI Automates Tasks, Not Entire Occupations
Labor Economists 40%Enterprise Management Researchers 35%Macroeconomic Policymakers 25%
ExplainerModel Compression
How Unstructured Pruning Removes 90% of a Model's Weights Without Accuracy Loss
Theoretical Researchers 40%Hardware Engineers 30%Deployment Practitioners 30%
ExplainerIn-Context Learning
The Mesa-Optimization Engine: How Transformers Simulate Gradient Descent to Learn In-Context
Theoretical Mechanists 40%Empirical Skeptics 35%AI Safety Researchers 25%
AI InfrastructureAcquisition Analysis
Nvidia Acquires Hugging Face for $12.9 Billion in Major AI Infrastructure Consolidation
Nvidia & Open-Source Advocates 45%Hardware Competitors & Antitrust Monitors 35%Enterprise Adopters 20%
ExplainerAI Evaluation
AllenAI's BenchMIRT Separates Raw Intelligence from Safety in LLM Evaluations
AI Safety Researchers 40%Enterprise Procurement Teams 30%Open-Source Evaluators 30%
ExplainerBrowser Inference
How 207 Open-Source WebGPU Kernels Turn a Desktop Browser Into a Local AI Inference Engine
WebAI Developers 40%Web Standards Bodies 35%Open-Source Maintainers 25%
AI Hardware Architecture
How Systolic Arrays Pass Data Through AI Chips to Execute Matrix Multiplication
Domain-Specific Architecture Proponents 40%General-Purpose Compute Advocates 40%Edge AI Researchers 20%
ExplainerCompute Infrastructure
How PagedAttention Partitions GPU Memory to Solve the LLM Context Bottleneck
Inference Engine Developers 60%Hardware Architects 40%
ExplainerModel Safety
How Outer Alignment Teaches an AI to Pass a Test While Inner Alignment Dictates Its True Goal
Mesa-Optimization Theorists 40%Empirical Deep Learning Researchers 35%Sociotechnical Evaluators 25%
AI BenchmarksPerformance Record
Nvidia's Nemotron-3-Ultra-CC Model Beats Top Human Score at International Informatics Olympiad
AI Researchers 40%Industry Observers 40%Competitive Programmers 20%
ExplainerModel Optimization
How Low-Rank Adaptation Compresses AI Fine-Tuning Into 0.01% of a Model's Parameters
Efficiency Researchers 40%Enterprise Deployers 35%Platform Providers 25%
ExplainerAgent Memory
How AI Agents Remember: Comparing Context Windows, RAG, and Episodic Storage
Retrieval & Efficiency Engineers 50%Long-Context Advocates 30%Enterprise Security & Governance 20%
ExplainerFederated Learning
How Federated Learning Trades Model Accuracy for Data Security
Privacy Advocates 35%AI Performance Researchers 35%Edge Hardware Engineers 30%
ExplainerPositional Bias
The Positional Advantage of the System Prompt: How Pre-pending Instructions to the Context Window Constrains LLM Output
Context Engineers 60%Architectural Determinists 40%
Research BriefPatent Law
How Global Patent Law Universally Rejects AI Inventors Over the Definition of "Conception"
Legal Traditionalists 40%Pro-AI Innovation Advocates 30%Transparency Proponents 30%
ExplainerNetwork Architecture
The Exponential Representational Advantage of Deep, Narrow Neural Networks Over Shallow, Wide Ones
Theoretical Computer Scientists 40%Applied AI Engineers 35%Hardware Architects 25%
ExplainerSLAM Algorithms
Computational Cost vs. Accuracy: Comparing EKF and Particle Filter Solutions for SLAM
Graph Optimization Advocates 50%Probabilistic Filtering Traditionalists 30%Algorithmic Analysts 20%
ExplainerAI Alignment
Bypassing the Reward Model: How Direct Preference Optimization Aligns Open-Source AI
Open-Source Developers 50%Alignment Theorists 30%Editorial Analysis 20%
ExplainerModel Decoding
How Temperature and Nucleus Sampling Shape Language Model Output
Deterministic Engineers 35%Creative Prompt Designers 35%Sampling Algorithm Researchers 30%
Autonomous AISecurity Policy
OpenAI Blocks Exploit Requests as GPT-6 Astra Achieves Autonomous Zero-Day Discovery
AI Developers 40%Enterprise Defenders 30%Security Researchers 30%
Research BriefNeural Network Architecture
How Residual Connections and ReLU Activation Prevent Vanishing Gradients in Deep Neural Networks
Applied AI Researchers 40%Theoretical Mathematicians 35%Systems Engineers 25%
ExplainerContinual Learning
How Elastic Weight Consolidation Stops Neural Networks From Forgetting
Biological AI Researchers 40%Compute-Constrained Engineers 35%Continual Learning Purists 25%
ExplainerExport Controls
Foreign Direct Product Rule Restricts Global AI Chips Made With US Tools
Export Control Advocates 40%Semiconductor Industry 35%Policy Skeptics 25%
ExplainerInference Optimization
How Speculative Decoding Accelerates Large Language Models Without Quality Loss
Hardware Engineers 35%Open-Source AI Community 35%Enterprise Cloud Providers 30%
Autonomous AISecurity Assessment
Booz Allen Identifies Anthropic's Mythos as First AI Capable of Autonomous Cyberattacks Amid Debate Over Attack Harnesses
Threat Intelligence Analysts 50%Systems Security Researchers 50%
Frontier ModelsLegislative Proposal
Sanders and Casar Introduce Legislation to Ban Artificial Superintelligence
Statutory Interventionists 35%Research Pragmatists 35%Industry Stakeholders 30%
ExplainerSemantic Search
Mapping Meaning: How Vector Embeddings Translate Human Language into AI Geometry
Commercial AI Providers 40%Open-Source Developers 30%Enterprise Infrastructure Teams 30%
Global AI PolicyDiplomatic Consensus
U.S. Advances 'Carolina Principles' at G20 to Promote Innovation-First AI Regulation
Innovation Advocates 40%Global Market Stabilizers 35%Precautionary Regulators 25%
ExplainerCompute Thresholds
US and EU AI Compute Thresholds: Translating FLOPs Into Hardware Requirements
Innovation Protectionists 40%Precautionary Regulators 35%Algorithmic Skeptics 25%
ExplainerAI Training Algorithms
Gradient Descent and Loss Functions: How Neural Networks Minimize Training Error
Applied Practitioners 45%Theoretical Purists 35%Hardware Efficiency Advocates 20%
ExplainerAI Self-Correction
AI Self-Correction: How Language Models Critique and Refine Their Own Reasoning
AI Alignment Researchers 40%Enterprise Implementers 40%Efficiency Advocates 20%
ExplainerModel Transparency
Mechanistic Interpretability: Mapping Neural Network Features and Circuits to Decode AI
Mechanistic Interpretability Researchers 40%Empirical AI Safety Teams 30%Deep Learning Theorists 30%
Open-Weight ModelsTechnical Release
Tencent Open-Sources 770-Billion Parameter Hy4 Preview Model for Enterprise AI
Open-Source Developers 40%Enterprise Infrastructure Providers 35%AI Performance Analysts 25%
ExplainerAlgorithmic Bias
AI Bias: Why Data Engineering Fails and Systemic Auditing Remains Unstandardized
Socio-Technical Standards Bodies 40%Clinical Equity Researchers 30%Consumer Policy Advocates 30%
ExplainerAI Architecture
Predicting LLM Performance Through Scaling Laws of Compute, Data, and Parameters
Scale Maximalists 40%Efficiency Optimizers 35%Architectural Skeptics 25%
ExplainerAI Strategy
How Proprietary Data Flywheels Turn User Feedback into Defensible AI Moats
Enterprise Incumbents 45%AI Challengers & Startups 35%Legal & Compliance Experts 20%
ExplainerAI Tool Use
How Language Models Use Structured JSON to Execute External Function Calls
Agentic Optimists 40%Security & Reliability Skeptics 35%Open-Source Developers 25%
Research BriefAI Privacy Tech
Combining Federated Learning and Differential Privacy for Legal AI Training Under GDPR
Regulatory Compliance Advocates 50%AI Fairness Researchers 50%
ExplainerAI Copyright Law
Whether AI Training on Copyrighted Data Legally Constitutes Transformative Fair Use
AI Developers & Researchers 35%Copyright Holders & Creators 35%Legal & Regulatory Analysts 30%
Fusion ControlResearch Milestone
AI Framework PACMAN Achieves Millisecond-Level Control of Fusion Plasma
Plasma Physicists 40%AI Control Engineers 40%Energy Infrastructure Analysts 20%
Medical AIDiagnostic Breakthrough
AI Tool Detects Hidden Heart Disease From Routine ECGs in Under Two Seconds
Cardiology Researchers 45%Clinical Practitioners 35%Health Tech Industry 20%
ExplainerAI Hardware
Why AI Chips Trade Flexibility for Efficiency Across GPUs, TPUs, and ASICs
General-Purpose Hardware Advocates 40%Custom Silicon Developers 40%Edge Computing Proponents 20%
AI InfrastructureCapacity Deal
Anthropic Signs $35 Billion Compute Deal with Lambda to Secure Nvidia Capacity
AI Infrastructure Providers 35%Frontier Model Developers 35%Financial & Market Analysts 30%
Deep DiveExplainable AI
Post-Hoc AI Explainers Approximate Black-Box Logic Rather Than Revealing It
Post-Hoc Pragmatists 40%Intrinsic Purists 35%Regulatory Compliance Officers 25%
Explore more topics
Per page