TLDR AI 2024-06-25

Continuing the acquisition spree, OpenAI buys more foundational technology for using the desktop.

OpenAI has reversed its previous restrictive policies, now allowing current and former employees to participate equally in annual tender offers for stock sales. This change addresses concerns about liquidity access, especially given OpenAI's high valuation, and eliminates provisions that allowed forced share repurchases.

Amazon is reportedly developing an AI chatbot called "Metis" to compete with OpenAI's ChatGPT, expected to be unveiled later this year. The chatbot, which will be accessible via web browser, will utilize a new foundational model and share infrastructure with "Remarkable Alexa," Amazon's upcoming paid-tier Alexa version.

A list of curated computer vision papers from the 2024 CVPR conference.

PE-Rank is a new approach to passage ranking that improves efficiency by using single passage embeddings for context compression.

Toucan recently released a set of new text-to-speech models that it has extended to work with every ISO-639-3 standard language.

TRL is a Hugging Face library for training transformers with reinforcement learning. This example allows you to also perform the same process for vision-based language models like LLaVA.

Implementing batch parallel KV cache in MLX leads to dramatic inference time speed up for synthetic data generation and model completions.

The Mixture of Attention (MoA) approach optimizes sparse attention in large language models by tailoring unique sparse attention configurations for different heads and layers.

This project introduces CDMask and CDMaskFormer, two new models for remote sensing change detection.

Popular developer Shawn "swyx" Wang talks about hidden opportunities for traditional software developers looking to transition into AI. Specifically, looking at building on top of current tools, using AI for summarization, and more.

Apple has integrated generative AI into its core applications, enhancing features like Safari summaries, Mail categorization, and Siri's functionality, rather than creating standalone AI products, showcasing the company's focus on privacy and user control.

This post shows how to run a small Llama language model in a font file.

Ability AI, a marketing startup, has raised a round to build its marketing agent technologies.

June is a local voice chatbot that combines the power of Ollama, Hugging Face Transformers, and the Coqui TTS Toolkit.