The Daily Index — Friday, August 30, 2024 · TLDR Index

HEADLINES & LAUNCHES

TLDR AI

TLDR AI

OPENAI AND ANTHROPIC WILL SHARE THEIR MODELS WITH THE US GOVERNMENT

OpenAI and Anthropic have agreed to allow the US government early access to their major new AI models before public release to enhance safety evaluations as part of a memorandum with the US AI Safety Institute.

theverge.com · 1 min

TLDR AI

THE MAMBA IN THE LLAMA: DISTILLING AND ACCELERATING HYBRID MODELS

Mambda models are hard to accelerate with speculative decoding because of the lack of a KV cache for backtracking. This paper, from some of the original authors, showcases some novel acceleration algorithms and distillation methods.

arxiv.org · 31 min

TLDR AI

EFFICIENT LLM SCHEDULING BY LEARNING TO RANK

When serving many concurrent requests to a large language model, things often get bottlenecked because we don't know how long output generation will be, so we get head of line bottlenecks. However, you can learn to rank relative lengths between requests and serve the shortest ones first, which leads to a 6.5x higher throughput for multi batch generation.

arxiv.org · 24 min

TLDR AI

MULTI-TASK SCENE UNDERSTANDING

MTMamba++ is a new model architecture designed to enhance multi-task scene understanding. This approach uses a Mamba-based decoder featuring two core blocks, STM and CTM, to capture long-range dependencies and improve cross-task interactions.

arxiv.org · 16 min

TLDR AI

HERMES FUNCTION CALLING DATASET (HUGGING FACE HUB)

Great open dataset from Nous Research for training models to call functions.

huggingface.co

TLDR AI

QWEN 2 VL

The Qwen team has released vision language models built on top of Qwen 2.

qwenlm.github.io · 12 min

TLDR AI

RAW IMAGE PROCESSING

RAW-Adapter is a new approach that adapts sRGB pre-trained models to handle camera RAW data effectively.

cuiziteng.github.io · 3 min

TLDR AI

LLAMA USAGE DOUBLE MAY THROUGH JULY

Meta has released some statistics about Llama model usage. It found strong demand for its models being deployed in enterprise settings.

ai.meta.com · 12 min

TLDR AI

3D MEDICAL IMAGE ANNOTATION

This study has adapted the Segment Anything Model 2 (SAM 2), originally designed for video annotation, to streamline the annotation of 3D medical images.

arxiv.org · 23 min

TLDR AI

ANTHROPIC SAYS IT IS CLOSER TO BACKING CALIFORNIA'S AI BILL AFTER LAWMAKERS MADE SOME FIXES

Anthropic has shifted to a cautious endorsement of California's AI regulation bill SB 1047 after key amendments. Changes include reduced penalties, clearer language, and increased reporting requirements. The bill aims to ensure safety in AI practices, gaining support from Anthropic yet facing opposition from OpenAI and Meta.

businessinsider.com · 3 min

TLDR AI

WISPR FLOW (PRODUCT LAUNCH)

Flow is a Mac dictation app that lets you speak naturally and writes in your style in every application, 3x faster than a keyboard.

flowvoice.ai

TLDR AI

OPENAI SAYS CHATGPT USAGE HAS DOUBLED SINCE LAST YEAR

OpenAI announced that ChatGPT has surpassed 200 million weekly active users, doubling its user base from a year ago, and noted that 92% of Fortune 500 companies are using its products.

links.tldrnewsletter.com · 1 min

TLDR AI

TIKTOK OWNER BYTEDANCE LAUNCHES NEW VIDEO SEARCH TOOL

ByteDance has introduced Douyin Search, an app for searching short video content on China's version of TikTok, directly competing with Baidu's search dominance.

scmp.com · 2 min