SMRTR AI

AI stories

Get hand-picked daily summaries of the best, most informative AI articles from around the web. Browse recent summaries, then jump to the original source when you want the full article.

Latest story• Dec 30, 2024

ts_zip: Text Compression using Large Language Models

The ts_zip utility uses a Large Language Model to compress text files more efficiently than conventional tools. It requires a GPU and 4GB RAM, processing up to 1 MB/s on an RTX 4090. While supporting various languages and source code, it's optimized for English texts. Compression is measured in bits per byte (bpb). For instance, ts_zip compressed the 100MB "enwik8" file to 13.8MB (1.106 bpb), surpassing xz's 24.9MB...

Read summary Original source

Latest stories

Introducing ZeroShape: A Strong Regression-Based Zero-Shot 3D Shape Reconstruction Method

ZeroShape, a new regression-based model for 3D shape reconstruction from single images, outperforms state-of-the-art methods while using less computing power and data. The researchers also created a large evaluation benchmark...

Read summary Original

Try Llama 3.1 8B in Your Browser: AQLM.rs Delivers Al at Your Fingertips

Llama 3.1 8B, an advanced language model, can now run directly in web browsers using WebAssembly and extreme compression techniques. The model is compressed to just 2.5 GB using 2-bit quantization, allowing it to outperform...

Read summary Original

How-to Use AI to See Your Data in 3D

TrustGraph's Data Workbench 3D visualizer enhances data analysis with immersive, interactive 3D visualization. This tool helps uncover hidden relationships, provides intuitive navigation, and offers enhanced understanding of...

Read summary Original

Did China land a sucker punch on Sam Altman’s OpenAI in AI scaling secret race?

A controversy has emerged over the origins of AI scaling laws, crucial for large language models. OpenAI's 2020 paper on scaling laws may have built on earlier work by Baidu. Former OpenAI VP Dario Amodei observed scaling effects...

Read summary Original

3 Big Things Going On At OpenAI, From Business Changes To AI Evolution

OpenAI, led by Sam Altman, is rapidly advancing artificial intelligence technology. ChatGPT is challenging Google's dominance in search, offering a more intuitive and conversational approach. The company is maintaining both...

Read summary Original

SMRTR AI

Get the next batch of curated stories in your inbox.

This archive is built from SMRTR newsletter stories. Subscribe for hand-picked stories without the extra noise.

More recent stories

iWallet Begins Offering Voice AI Payments

iWallet is introducing voice AI technology for payments, becoming the first to offer a PCI-compliant telephone order system using voice instead of manual entry. This innovation aims to enhance customer experience, improve...

Read summary Original

How I run LLMs locally

A tech enthusiast describes their setup for running large language models locally on a high-end PC, covering hardware needs, software tools, model selection, and the advantages of local LLM use such as data control and reduced...

Read summary Original

Can LLMs Accurately Recall the Bible

Large language models (LLMs) vary in their ability to accurately recall Bible verses. A benchmark test evaluated 11 different LLMs across six scenarios, from popular to obscure verses. Larger models like GPT-4 and Claude 3.5...

Read summary Original

Building Life With Generative AI

Generative AI is being applied to life sciences, enabling precise protein engineering and antibody discovery. Flagship Pioneering's Geoff von Maltzahn explains how AI models can create improved antibodies for diseases like cancer...

Read summary Original

Geoffrey Hinton Predicts Human Extinction At The Hands Of AI. Here’s How To Stop It

AI experts warn of potential extinction risks within 30 years, advocating for regulation, global cooperation, and education focused on adaptability and ethics to complement AI while safeguarding humanity's future.

Read summary Original

5 Ways AI Is Keeping The Booming Air Travel Industry On Track

AI is transforming air travel by improving efficiency, reducing delays, and enhancing safety. It diagnoses issues, predicts disruptions, optimizes operations, boosts sustainability, and improves passenger experiences, potentially...

Read summary Original

AI data centers could be ‘distorting’ the US power grid

Data center growth for AI computing is causing power grid concerns, with reports of nearby households experiencing severe power distortions that could damage appliances, increase fire risk, and cause outages, though some...

Read summary Original

QVQ: To See the World with Wisdom

The problem involves finding the derivative of a product function h(x) = f(x) * g(x) at x = 2, using given values for f(x), g(x), and their derivatives. Using the product rule for derivatives, h'(x) = f'(x)*g(x) + f(x)*g'(x). By...

Read summary Original

Five Apple Intelligence Features Coming in 2025

Apple plans to release major Apple Intelligence updates in 2025, including Priority Notifications, enhanced Siri capabilities, and expanded language support. New features will allow Siri to understand on-screen context, integrate...

Read summary Original

The Structure of Neural Embeddings

Deep neural networks produce structured embeddings with hierarchical organization, linear feature representation, and superposition of concepts. These latent spaces exhibit properties like manifold structure, universality across...

Read summary Original

You can now rent Google's most powerful AI chip: Trillium TPU underpins Gemini 2.0 and will put AMD and Nvidia on high alert

Google's sixth-gen AI chip, Trillium, is now available for rent, offering 4x the training performance of previous versions. Used to train Google's Gemini 2.0 AI model, it boasts 2.5x improvement in training performance per...

Read summary Original

Why DeepSeek’s new AI model thinks it’s ChatGPT

DeepSeek, a Chinese AI lab, released DeepSeek V3, an AI model outperforming many competitors on benchmarks. However, the model often misidentifies itself as ChatGPT or GPT-4, suggesting it may have been trained on datasets...

Read summary Original

AI data centers are becoming 'mind-blowingly large'

Data centers for AI are expanding dramatically, with some reaching over two kilometers in length and multiple stories high. This growth is driven by AI's increasing power needs, with some new facilities requiring a gigawatt to...

Read summary Original

DeepSeek v3 – A 671B parameter AI Language Model

DeepSeek v3, a groundbreaking AI language model, boasts 671 billion total parameters with 37 billion activated per token. This advanced Mixture-of-Experts model, trained on 14.8 trillion tokens, achieves state-of-the-art...

Read summary Original

OpenAI announces plan to transform into a for-profit company

OpenAI will transition to a for-profit Public Benefit Corporation in 2025, balancing AI development funding with nonprofit activities, despite potential legal challenges from tech leaders Elon Musk and Mark Zuckerberg.

Read summary Original

Turning The Tide: Generative AI’s Role In Future Tsunami Management

AI enhances disaster management by predicting events, guiding rescuers, assessing damage, and improving infrastructure, using data analysis to save lives and boost preparedness for future catastrophes.

Read summary Original

Breaking up is hard to do: Chunking in RAG applications

Retrieval-augmented generation (RAG) is increasingly used to ground LLM responses in specific data sources. Proper data chunking is crucial for effective RAG systems, as chunk size and method significantly impact search results...

Read summary Original

'$100B in profits': Microsoft, OpenAI's secret definition of AGI revealed

OpenAI and Microsoft have defined AGI as a system capable of generating $100 billion in profits, as part of a 2023 agreement. OpenAI's non-profit board will determine when AGI is achieved. The company projects losses of $44...

Read summary Original

Back to all archives Load more stories