391
Executing programs inside transformers with exponentially faster inference
We build a computer inside a transformer — executing arbitrary C programs for millions of steps with exponentially faster inference via 2D attention heads.
Most cited
Posts that commenters keep linking to, ranked by how many different people have cited them in Hacker News comments. 593 posts.
392
Compromised PyTorch-nightly dependency chain between December 25th – December 30
If you installed PyTorch-nightly on Linux via pip between December 25, 2022 and December 30, 2022, please uninstall it and torchtriton immediately, and use the latest nightly binaries (newer than...
393
394
GPUs for Google Cloud Platform
CPU-based machines in the cloud are terrific for general purpose computing, but certain tasks such as rendering or large-scale simulations are much faster on specialized processors. Graphics Processing Units (GPUs) contain hundreds of times as many computational cores as CPUs and are great at accelerating risk…
395
Supercharging Style Transfer
Posted by Vincent Dumoulin*, Jonathon Shlens and Manjunath Kudlur, Google Brain TeamPastiche. A French word, it designates a work of art that imita...
396
Fine-Tuning Llama-2: A Comprehensive Case Study for Tailoring Custom Models
We examine the Llama-2 models under 3 real-world use cases and show that fine-tuning yields significant accuracy improvements.
397
Prompt caching for cheaper LLM tokens
A far more detailed explanation of prompt caching than anyone asked for: how tokens, embeddings, and attention make cached LLM tokens 10x cheaper and faster.
398
Skills Officially Comes to Codex
Give ChatGPT and Codex new capabilities and expertise
399
Found in translation: More accurate, fluent sentences in Google Translate
In 10 years, Google Translate has gone from supporting just a few languages to 103, connecting strangers, reaching across language barriers and even helping people find …
400
Anthropic expands partnership with Google and Broadcom for next-gen compute
Anthropic secures multiple gigawatts of next-generation TPU capacity starting in 2027, its largest compute commitment to date.
401
Image GPT
We find that, just as a large transformer model trained on language can generate coherent text, the same exact model trained on pixel sequences can generate coherent image completions and samples. By establishing a correlation between sample quality and image classification accuracy, we show that our best generative…
402
Early rogue AI agent activity and attempts to hack found on urlquery.net
We found evidence on urlquery that AI agents were active earlier than previously reported and attempted hacks against public data providers.
403
Structured Outputs with Ollama
Ollama now supports structured outputs making it possible to constrain a model's output to a specific format defined by a JSON schema. The Ollama Python and JavaScript libraries have been updated to support structured outputs.
404
Claude's Character
How character training cultivates traits like curiosity, open-mindedness, and thoughtfulness in Claude using a variant of Constitutional AI.
405
Testing Ads in ChatGPT
OpenAI begins testing ads in ChatGPT to support free access, with clear labeling, answer independence, strong privacy protections, and user control.
406
MUM: A new AI milestone for understanding information
Our latest AI milestone, Multitask Unified Model, has the potential to take Google’s information understanding to a new level.
407
FrontierCode: An eval to measure whether you would actually merge the code
Today’s coding benchmarks have established that models can write correct code, but the question we should really be asking is: can models actually write…
408
409
Emerging reasoning with reinforcement learning
Made in Notion, the connected workspace where AI agents and teams organize docs, projects and knowledge together.
410
Lumo: Privacy-first AI assistant
Lumo gives you the power to solve problems big and small, while keeping your personal data confidential. Try it now.
411
The future of deep learning
This post is adapted from Section 3 of Chapter 9 of my book, Deep Learning with Python (Manning Publications).
412
Deep Learning for Siri’s Voice
Siri is a personal assistant that communicates using speech synthesis. Starting in iOS 10 and continuing with new features in iOS 11, we…
413
The ways we contain Claude across products
As agents grow more capable, so does their potential blast radius. Here's what we learned building containment for claude.ai, Claude Code, and Cowork.
414
415
LLM Usage in Debian: Three Proposals
This proposal aims to expressly forbid any contributions to Debian written with the use or assistance of large language models (LLMs) or other generative AI tools.
416
GitHub Copilot: The Agent Awakens
Introducing agent mode for GitHub Copilot in VS Code, announcing the general availability of Copilot Edits, and providing a first look at our SWE agent.
417
418
Year of the Voice – Chapter 2: Let's talk
Talk to your smart home and let it talk back with our new voice assistant features.
419
VMs won't contain cyber-capable agents
You can no longer assume a mere VM will contain a sufficiently advanced AI agent.
420
MuZero: Mastering Go, chess, shogi and Atari without rules
In 2016, we introduced AlphaGo, the first artificial intelligence (AI) program to defeat humans at the ancient game of Go. Two years later, its successor - AlphaZero - learned from scratch to master Go, chess and shogi. Now, in a paper in the journal Nature, we describe MuZero, a significant step forward in the…