Explained · Page 3
guides, explainers and opinion for the AI-curious
Sentence Transformers Gets a Smarter Search Engine Under the Hood
A new update to one of the most popular AI search libraries adds a technique called late interaction, which matches text word by word instead of squeezing everything into a single number. Here is what that means and why it matters.
AI Models May Not Need to Forget Everything: The Case for Smarter Unlearning
New research from Apple ML Research suggests some training data has so little effect on an AI model that removing it is a waste of time and money.
Same AI Memory, Fewer Tokens: How a New Approach Beats a Leading Agent System at a Fraction of the Cost
Two AI systems both teach agents to learn from their own mistakes. One sends every lesson every time. The other sends only what each model can actually use. The difference shows up in the bill.
A New Trick Slashes the Cost of Shrinking Giant AI Models
Researchers have found a way to train smaller, cheaper AI models from giant ones using a fraction of the memory previously required, opening the door to experiments that once needed a warehouse of hardware.
Apple's AI Shortcut: How a 'Draft and Check' Trick Makes Reasoning Models Twice as Fast
Apple ML Research has built a smarter way to speed up AI thinking, one that checks meaning instead of counting exact words. It could cut the cost of running powerful AI in half.
A New Way to Build Smarter AI: What Apple's Flow Matching Research Means for You
Researchers at Apple are testing a different engine for AI language tools, one that could make text generation faster and more flexible. Here is what it is and why it matters.
New AI Tool Tackles Tricky Questions: DEEPAMBIGQAGEN Unveiled
Apple ML Research introduces DEEPAMBIGQAGEN, a tool designed to boost AI's ability to handle complex questions.
Apple Researchers Find a Hidden Glitch That Can Quietly Corrupt AI-Generated Images
A small number of rogue data points, called outlier tokens, can throw off the systems that create AI images. New research explains what they are and how to fix them.
The Power Grid Is Struggling. A New AI Course Wants to Train the Engineers Who Can Fix It
America's electrical grid was designed for a calmer era. Now AI tools are being drafted to keep the lights on, and a new online programme is teaching engineers how to use them.
353,000 signed up. 6,000 shipped. Inside Google's 'vibe coding' bootcamp
Google and Kaggle's five-day crash course pulled in a third of a million learners to build AI agents by chatting to their computer. Here's what actually happened.
GraphRAG vs. plain RAG: the honest scorecard
A knowledge graph can make AI answers far better, but only for certain questions, and the indexing bill can shock you. Here is what five studies actually found.
Why AI Image Models Still Make Things Up, and What Apple's Researchers Are Doing About It
A new study from Apple ML Research digs into why multimodal AI models hallucinate, meaning they describe images with confident-sounding details that simply are not there, and how a training technique called preference alignment could fix it.
Two-thirds of people think ChatGPT is conscious. A philosopher says we've been fooling ourselves for decades.
A new survey finds most people attribute some form of feeling to AI chatbots. One scholar argues we've had the wrong idea about minds and bodies long before the first chatbot appeared.
Every major AI model leans left on politics, even Grok, half the time
A small research lab ran 16 leading AI models through a well-known political quiz thousands of times. All but one landed consistently in the libertarian-left zone. The one exception was Elon Musk's Grok, which couldn't make up its mind.
Why AI Vision Systems Miss What's Right in Front of Them
Apple ML Research has a new tool that finds the hidden patterns behind AI mistakes in object detection, and its findings matter for anyone relying on AI to spot things in the real world.