Hacker News headlines

3

Scaling Laws vs. Model Architectures: How Does Inductive Bias Influence Scaling? (arxiv.org)

4 hours ago | georgehill | arxiv.org | newest

2

The "it" in AI models is the dataset (nonint.com)

22 hours ago | georgehill | nonint.com | newest

15

Snowflake Arctic: LLM for Enterprise AI — Efficiently Intelligent, Truly Open (snowflake.com)

a day ago | georgehill | snowflake.com | newest

2

OpenAI: Training LLMs to Prioritize Privileged Instructions (arxiv.org)

a day ago | georgehill | arxiv.org | newest

2

The Illusion of State in State-Space Models (arxiv.org)

3 days ago | georgehill | arxiv.org | newest

6

LLM Leaderboard with explanations of what each score means (stanford.edu)

4 days ago | georgehill | stanford.edu | frontpage

239

[flagged] Why you should not apply to YC (twitter.com/dvassallo)

3 days ago | georgehill | twitter.com | best

0

Naval: Attempting to Define AGI (air.chat)

6 days ago | georgehill | air.chat | newest

1

Aide: The Machine Learning CodeGen Agent (github.com/wecoai)

a week ago | georgehill | github.com | newest

2

Clear Eyes, Full Heart (2lr.substack.com)

a week ago | georgehill | substack.com | newest

2

Mechanics of Next Token Prediction with Self-Attention (arxiv.org)

a week ago | georgehill | arxiv.org | frontpage

2

JetMoE: Reaching Llama2 Performance with 0.1M Dollars (arxiv.org)

a week ago | georgehill | arxiv.org | newest

1

From Words to Numbers: Your Large Language Model Is a Capable Regressor (arxiv.org)

a week ago | georgehill | arxiv.org | newest

1

Language network as a natural kind within the landscape of the human brain (nature.com)

a week ago | georgehill | nature.com | newest

2

Liquid AI's Ramin Hasani on liquid neural networks [video] (youtube.com)

a week ago | georgehill | youtube.com | newest

2

The Brain Forming a Memory (twitter.com/brianroemmele)

2 weeks ago | georgehill | twitter.com | newest

1

An Open-Source AI-Powered Answer Engine with a Generative User Interface (morphic.sh)

2 weeks ago | georgehill | morphic.sh | newest

1

How to Fix Legal Immigration in America [video] (youtube.com)

2 weeks ago | georgehill | youtube.com | newest

2

Securing Canada's AI Advantage (pm.gc.ca)

2 weeks ago | georgehill | pm.gc.ca | newest

2

Google lost ground in the AI race (ft.com)

2 weeks ago | georgehill | ft.com | newest

2

OpenAI: Testing Usage-Based GPT Earnings with US Builders (twitter.com/openai)

4 weeks ago | georgehill | twitter.com | newest

2

Collaborate with point cloud for manufacturing sites (samp.ai)

a month ago | georgehill | samp.ai | newest

3

Revealing OpenAI's plan to create AGI by 2027 [PDF] (drive.google.com)

a month ago | georgehill | google.com | newest

1

Rename Your Images and Videos with AI (keepitshot.com)

2 months ago | georgehill | keepitshot.com | newest

2

Ask HN: How does plagarism checker work?

2 months ago | georgehill | ycombinator.com | newest

3

Software-Defined Tensor Streaming Multiprocessor for Large-Scale ML [pdf] (groq.com)

2 months ago | georgehill | groq.com | newest

27

Grand-Master Level Chess Without Search: Modeling Choices and Their Implications (gist.github.com)

2 months ago | georgehill | github.com | best

2

Carta launches a startup shutdown service (henrysward.medium.com)

2 months ago | georgehill | medium.com | frontpage

7

Convert 'Screenshot.png' to 'Appropriate name.png' Using GPT-4 Vision (keepitshot.com)

3 months ago | georgehill | keepitshot.com | newest

0

Training Neural Networks Is NP-Hard in Fixed Dimension (arxiv.org)

3 months ago | georgehill | arxiv.org | newest

1

Satoshi: Hello World (twitter.com/satoshi)

3 months ago | georgehill | twitter.com | newest

96

Some updates on our AI efforts (threads.net)

3 months ago | georgehill | threads.net | best

94

Bing Gained Less Than 1% Market Share Since Adding Bing Chat (seroundtable.com)

3 months ago | georgehill | seroundtable.com | best

1

Local LLM movement feels like early days of PCs vs. Mainframes (reddit.com)

3 months ago | georgehill | reddit.com | newest

25

Phi-2: Self-Extend Boosts Performance, Extends Context to 8k Without Training (reddit.com)

3 months ago | georgehill | reddit.com | best

1

Google confirms it just laid off around a thousand employees (theverge.com)

3 months ago | georgehill | theverge.com | newest

3

Serving Open Source Models 4x faster than vLLM by quantizing with ~no tradeoffs (fireworks.ai)

3 months ago | georgehill | fireworks.ai | newest

1

LNX: Using Tantivy to Build One of the Fastest Search Engines Around [video] (youtube.com)

3 months ago | georgehill | youtube.com | newest

1

Moe-Mamba: Efficient Selective State Space Models with Mixture of Experts (arxiv.org)

3 months ago | georgehill | arxiv.org | newest

3

Evaluating LLMs with CommonGen-Lite (github.com/allenai)

3 months ago | georgehill | github.com | newest

65

Turing Complete Transformers: Two Transformers Are More Powerful Than One (openreview.net)

3 months ago | georgehill | openreview.net | best

6

Human-Like Browser Automation with GPT-4 Vision and MeiliSearch (github.com/vignshwarar)

3 months ago | georgehill | github.com | frontpage

1

Create browser automation as if you were teaching a human using GPT-4 Vision (github.com/vignshwarar)

3 months ago | georgehill | github.com | newest

6

Ask HN: What's the name of the website where the structure changes as I scroll?

3 months ago | georgehill | ycombinator.com | newest

0

The essay "How to Start a Startup," which grew into Y Combinator (twitter.com/paulg)

4 months ago | georgehill | twitter.com | newest

2

Ask HN: Video Understanding

4 months ago | georgehill | ycombinator.com | frontpage

1

GPT-Vision first most reliable open-source browser automation (github.com/vignshwarar)

4 months ago | georgehill | github.com | newest

2

Prompting-Based Methods for Text Ranking Using Large Language Models (reachsumit.com)

4 months ago | georgehill | reachsumit.com | newest

3

Bard. Is. Insanely. Useful (twitter.com/cgarciae88)

4 months ago | georgehill | twitter.com | newest

4

Unfortunately, No AGI Yet (twitter.com/omarsar0)

4 months ago | georgehill | twitter.com | newest

1

DeWave: Discrete EEG Waves Encoding for Brain Dynamics to Text Translation (openreview.net)

4 months ago | georgehill | openreview.net | newest

31

PowerInfer: Fast Large Language Model Serving with a Consumer-Grade GPU [pdf] (sjtu.edu.cn)

4 months ago | georgehill | sjtu.edu.cn | best

1

Cargo make: Rust task runner and build tool (github.com/sagiegurari)

4 months ago | georgehill | github.com | newest

2

Emmett Shear on AI's culture wars (mercury.com)

4 months ago | georgehill | mercury.com | newest

2

xAI: A Spectral Condition for Feature Learning (arxiv.org)

4 months ago | georgehill | arxiv.org | newest

17

A Mathematical Perspective on Transformers (arxiv.org)

4 months ago | georgehill | arxiv.org | best

2

Expanded legal protections and improvements to our API (anthropic.com)

4 months ago | georgehill | anthropic.com | newest

1

Harvey Raises $80M Series B from Elad GIL, Kleiner Perkins, OpenAI and Sequoia (harvey.ai)

4 months ago | georgehill | harvey.ai | newest

47

Mapping the semantic void: Strange goings-on in GPT embedding spaces (lesswrong.com)

4 months ago | georgehill | lesswrong.com | best

166

Word2Vec received 'strong reject' four times at ICLR2013 (openreview.net)

4 months ago | georgehill | openreview.net | best

3

Point Transformer V3: Simpler, Faster, Stronger (arxiv.org)

4 months ago | georgehill | arxiv.org | newest

70

OpenAI employee: GPT-4.5 rumor was a hallucination (twitter.com/willdepue)

4 months ago | georgehill | twitter.com | best

1

Agent Attention: On the Integration of Softmax and Linear Attention (arxiv.org)

4 months ago | georgehill | arxiv.org | newest

2

Agent Attention: On the Integration of Softmax and Linear Attention (github.com/leaplabthu)

4 months ago | georgehill | github.com | frontpage

1

My First $1 Online (marclou.beehiiv.com)

4 months ago | georgehill | beehiiv.com | newest

1

NeurIPS Test of Time: Tomas Mikolov's Insights on Word2vec, GloVe, and AI (pastebin.com)

4 months ago | georgehill | pastebin.com | newest

1

Mistral: $0.00 / 1M output tokens (openrouter.ai)

4 months ago | georgehill | openrouter.ai | newest

0

Infinite Inference Power for AI [video] (youtube.com)

4 months ago | georgehill | youtube.com | newest

1

SwitchHead: Accelerating Transformers with Mixture-of-Experts Attention (arxiv.org)

4 months ago | georgehill | arxiv.org | newest

0

MLC LLM: Universal Language Model Deployment Across Diverse Hardware and Apps (mlc.ai)

4 months ago | georgehill | mlc.ai | newest

1

Advance State of the Art Model and Tooling Support in Azure AI Studio (microsoft.com)

4 months ago | georgehill | microsoft.com | newest

3

Solve Puzzles. Learn CUDA (github.com/srush)

4 months ago | georgehill | github.com | newest

2

StatQuest: Word Embedding with PyTorch and Lightning (lightning.ai)

4 months ago | georgehill | lightning.ai | newest

2

Universal Approximators: State-Space Models with Memory Decay (arxiv.org)

4 months ago | georgehill | arxiv.org | newest

1

Anyone hear of GPT4.5 drop today? (reddit.com)

4 months ago | georgehill | reddit.com | newest

3

Balaji Fund (angellist.com)

4 months ago | georgehill | angellist.com | newest

34

[flagged] Kanye West Launches Yews News Platform (yews.news)

4 months ago | georgehill | yews.news | newest

6

Elon Musk Is Planning a New University in Austin (bloomberg.com)

4 months ago | georgehill | bloomberg.com | newest

1

Mistral-Medium vs. GPT-4's code generation (twitter.com/deliprao)

4 months ago | georgehill | twitter.com | newest

1

Twitter XSS + CSRF Vulnerability Exposes Account Takeover Risk (twitter.com/shoucccc)

4 months ago | georgehill | twitter.com | newest

2

Our Latest Experimental Tools and Technology (labs.google)

4 months ago | georgehill | labs.google | newest

37

Claude for Google Sheets (anthropic.com)

4 months ago | georgehill | anthropic.com | frontpage

1

Launched OpenAI 8 years and 1 day ago (twitter.com/gdb)

4 months ago | georgehill | twitter.com | newest

2

A Neural Corpus Indexer for Document Retrieval (arxiv.org)

4 months ago | georgehill | arxiv.org | newest

2

LLM360: Towards Transparent Open-Source LLMs (arxiv.org)

4 months ago | georgehill | arxiv.org | newest

3

EdgeSAM: Prompt-in-the-Loop Distillation for On-Device Deployment of Sam (arxiv.org)

4 months ago | georgehill | arxiv.org | newest

2

Together Inference Engine – the fastest inference available (together.ai)

4 months ago | georgehill | together.ai | newest

1

Also how GPT-4 works (youtu.be)

4 months ago | georgehill | youtu.be | newest

134

GigaGPT: GPT-3 sized models in 565 lines of code (cerebras.net)

4 months ago | georgehill | cerebras.net | best

2

Official PR Reveals the Inference Code for Mixtral 8x7B (github.com/vllm-project)

4 months ago | georgehill | github.com | newest

149

Mistral: Our first AI endpoints are available in early access (mistral.ai)

4 months ago | georgehill | mistral.ai | best

239

Mixtral of experts (mistral.ai)

4 months ago | georgehill | mistral.ai | best

2

LLMs as Copilots for Theorem Proving in Lean (github.com/lean-dojo)

4 months ago | georgehill | github.com | newest

1

Combining Continuous-Time, Recurrent, and Convolutional Models (stanford.edu)

4 months ago | georgehill | stanford.edu | newest

1

Image to Text ROS2 (gist.github.com)

4 months ago | georgehill | github.com | newest

41

Mathematicians have found a new upper limit to the Ramsey number (quantamagazine.org)

4 months ago | georgehill | quantamagazine.org | best

1

Play with mixtral-8x7b (vercel.ai)

4 months ago | georgehill | vercel.ai | newest

2

Automating Continual Learning (arxiv.org)

4 months ago | georgehill | arxiv.org | frontpage

1

The EU AI Act negotiations ended (twitter.com/ylecun)

4 months ago | georgehill | twitter.com | newest

2

OpenAI's Altman Ouster Was Result of Drawn-Out Tensions (bloomberg.com)

4 months ago | georgehill | bloomberg.com | newest