Machine learning
Source reports covering Machine learning. Stories will appear when analysis and community reactions are ready.
66 reports · page 2 of 2
Connected topics are temporarily unavailable. You can still browse stories below.
Source reports
Speculative Speculative Decoding (SSD)
research
2 min read
Speculative decoding accelerates autoregressive inference by using a fast draft model to predict upcoming tokens from a slower target model. It verifies predictions in parallel with a single forward pass of the target model, addressing the sequential dependency bottleneck.
This musician built an AI clone of her voice so anyone can sing as her
tool
4 min read
Holly Herndon has developed an AI voice clone that allows users to create music using her custom models. Her journey into machine learning began in 2015, evolving from initial "scratchy" outputs to sophisticated tools for musical expression.
Decision trees – the unreasonable power of nested decision rules
research
6 min read
Decision Trees create sequential rules that split data into distinct regions for classification. Entropy is used to measure information and identify regions with significant data separation.
10-202: Introduction to Modern AI (CMU)
tool
5 min read
The course "10-202: Introduction to Modern AI" covers the workings of modern AI systems, focusing on machine learning methods and large language models (LLMs) such as ChatGPT, Gemini, and Claude. The curriculum emphasizes the contemporary understanding of AI, primarily relating to chatbot technologies used daily.
Building a Minimal Transformer for 10-digit Addition
tool
1 min read
A minimal transformer model has been developed to perform 10-digit addition tasks. The model demonstrates the ability to learn and execute arithmetic operations effectively.
Looks like it is happening
research
2 min read
Data from December 2022 to December 2025 shows a steady increase in submissions, with numbers rising from 800 in 2022 to 855 in 2025. From January 1 to February 15, 2026, submissions reached 617, indicating a year-over-year growth trend.
AI Reveals Unexpected New Physics in the Fourth State of Matter
research
8 min read
Researchers at Emory University have used a machine learning technique to uncover unexpected features of non-reciprocal forces in many-body systems. The study combines a neural network with laboratory measurements from a dusty plasma, enhancing understanding of the fourth state of matter.
Music Discovery
tool
1 min read
AI-powered record discovery. Describe a vibe, name a record you love, or tell us how you're feeling — the clerk will dig through the crates for you.
Fast KV Compaction via Attention Matching
research
2 min read
Fast KV Compaction via Attention Matching addresses the limitations of key-value cache size in scaling language models for long contexts. It proposes a method that improves context management without the lossy effects of traditional summarization techniques.
Bridging Elixir and Python with Oban
tool
5 min read
Oban facilitates integration between Elixir and Python, allowing Elixir applications to utilize Python's advanced functionalities such as machine learning models and multimedia processing tools. This integration supports collaboration across teams and aids in gradual migration between programming languages.
As AI enters the operating room, reports arise of botched surgeries
news
13 min read
Johnson & Johnson's Acclarent added AI capabilities to its TruDi Navigation System for chronic sinusitis surgeries, utilizing a machine-learning algorithm to assist specialists. Reports have emerged of botched surgeries and misidentified body parts associated with this technology.
Reinforcement Learning from Human Feedback
research
2 min read
Reinforcement learning from human feedback (RLHF) is a key technique for deploying advanced machine learning systems. A new book provides an introduction to the core methods of RLHF for readers with a quantitative background.
Attention at Constant Cost per Token via Symmetry-Aware Taylor Approximation
2 min read
Self-attention mechanisms in Transformers typically incur costs that increase with context length, leading to higher demands for storage, compute, and energy. A new method using symmetry-aware Taylor approximation aims to maintain constant cost per token for self-attention, potentially alleviating these resource demands.
FlashAttention-T: Towards Tensorized Attention
6 min read
FlashAttention-T introduces a fully tensorized attention mechanism that leverages tensor-vector parallelism to enhance performance. This innovation aims to improve the efficiency of attention-based models in various applications.
Porting 100k lines from TypeScript to Rust using Claude Code in a month
16 min read
Vjeux aims to port 100,000 lines of code from TypeScript to Rust using Claude Code within a month. The project seeks to combine AI and algorithms to enhance codebase rewriting efficiency, inspired by large-scale programming challenges.
New UCLA AI tool targets Alzheimer's cases often missed in early diagnosis
2 min read
UCLA researchers are developing an AI tool to identify individuals at risk of Alzheimer's disease, aiming to improve early diagnosis and reduce disparities in affected communities. Early detection is crucial as new therapies can be prescribed for patients with early-stage Alzheimer's.