VLC tops 6B downloads, previews AI-generated subtitles

VLC media player has topped 6 billion downloads worldwide and is introducing an AI-powered subtitle system that can automatically generate real-time subtitles for any video using open-source AI models. The new feature, which can also translate subtitles into multiple languages, runs locally on users' devices, eliminating the need for internet connectivity or cloud services.

Read more

41% of Employers Worldwide Say They'll Reduce Staff by 2030 Due to AI

A recent survey by the World Economic Forum found that 41% of employers worldwide expect to reduce their staff by 2030 due to the increasing use of artificial intelligence, with jobs such as graphic designers and legal secretaries being particularly at risk. However, the survey also predicts a net growth in the number of jobs created over the next five years, with 170 million new jobs being created, although 92 million jobs will be displaced, resulting in a net growth of 78 million jobs.

Read more

Zuckerberg Approved AI Training on Pirated Books, Filings Say

Here is a 2-sentence summary of the text:

Mark Zuckerberg, CEO of Meta Platforms Inc., approved the use of a pirated book dataset to train the company's AI model LLaMA, according to unredacted court filings from a lawsuit alleging copyright infringement. The filings also reveal that Meta employees removed copyright information from the dataset, which was sourced from a controversial "shadow library" called LibGen, in an effort to conceal widespread copyright infringement.

Read more

41% of companies worldwide plan to reduce workforces by 2030 due to AI

41% of companies worldwide plan to reduce their workforces by 2030 due to the increasing use of artificial intelligence, which is expected to automate certain tasks. Meanwhile, 77% of employers intend to reskill and upskill their existing workers to better work alongside AI, as the technology continues to reshape the labor market and drive demand for specialist roles.

Read more

VLC player demos real-time AI subtitling for videos

The popular VLC video player has been demonstrated with a new feature that uses AI to generate subtitles for videos in real-time, with support for over 100 languages, all processed locally and offline. The feature, which is still in development, uses open-source AI models and can translate subtitles into the user's language, with no release date announced yet.

Read more

RAG with Differential Privacy

Retrieval-Augmented Generation (RAG) improves the quality of Large Language Models by providing fresh context, but it raises significant privacy concerns due to the risk of exposing confidential data. This paper proposes a solution to this problem by using differentially private token generation, a viable approach to private RAG that mitigates the risk of privacy breaches.

ArXiv

Agent Laboratory: Using LLM Agents as Research Assistants

Agent Laboratory is an autonomous framework that uses large language models to complete the research process, from literature review to report writing, and can produce high-quality research outputs with significant cost savings. The framework, which allows for human feedback and guidance, has been shown to generate state-of-the-art results, reduce research expenses by 84%, and improve overall research quality, potentially accelerating scientific discovery.

ArXiv

rStar-Math: Small LLMs Can Master Math Reasoning with Self-Evolved Deep Thinking

rStar-Math is a system that enables small language models to achieve state-of-the-art math reasoning capabilities, rivaling or surpassing larger models like OpenAI o1, through the use of Monte Carlo Tree Search and innovative training methods. The system has demonstrated impressive results, improving the performance of smaller models on math benchmarks and even solving a significant portion of problems from the USA Math Olympiad, comparable to the abilities of top high school math students.

ArXiv

Evaluation of Bfloat16, Posit, and Takum Arithmetics in Sparse Linear Solvers

The performance of alternative number formats, such as bfloat16, posit, and takum, is evaluated in the context of solving sparse linear systems using established solvers like LU, QR, and GMRES. The results show that tapered-precision posit and takum formats can achieve better accuracy and stability, particularly with takum arithmetic exhibiting exceptional stability even at low precision, in a wide range of real-world matrices.

ArXiv

Searching Latent Program Spaces

Program synthesis methods aim to generate programs that explain given input-output pairs, with recent neural network-based approaches learning program structures to narrow the search space. The proposed Latent Program Network (LPN) algorithm learns a distribution over latent programs, enabling efficient search and test-time adaptation, and outperforms other algorithms by generalizing beyond its training distribution and adapting to unseen tasks.

ArXiv

Open Agents – Secure, Composable AI Agents

The OpenAgentSpec is a standard for defining and interacting with generative AI agents, aiming to provide a secure and consistent framework for their development and deployment. By establishing a formal standard, OpenAgentSpec seeks to democratize access to AI, enable secure and autonomous agent interactions, and facilitate collaboration between agents across different organizations and boundaries.

Repo

Show HN: AgentScript AI – Build Agents that think in code

AgentScript is an open-source framework for building AI agents that think in code, allowing language models (LLMs) to generate executable code that can be run in a dedicated runtime with resumability, state persistence, and interactivity. The framework enables LLMs to express execution plans as code, making it possible for agents to think more abstractly about tasks and work with dynamic data without needing to know all the details upfront.

Repo

Make Llama 3.1 8B talk in Rick Sanchez's style

This project, called Rick LLM, aims to make the Llama 3.1 8B language model speak like Rick Sanchez from the animated series Rick and Morty by creating a custom dataset from transcripts, fine-tuning the model using Unsloth's optimizations on Lambda Labs GPUs, and deploying it to Ollama for local use. The project is divided into three main parts: dataset creation, model fine-tuning, and model deployment, with detailed instructions and code provided for each step.

Repo

Show HN: pay-respects – RIP command errors and keep yourself in the flow

Pay Respects is a tool that suggests fixes to incorrect console commands by pressing F, offering blazing fast and accurate suggestions with support for AI integration and modular customization. It can be installed through various package managers, pre-built binaries, or compiled from source using Cargo, with configuration options available for environment variables and AI settings.

Repo

Show HN: Experiment with DSPy optimzers, track performance of LLM-features

LangWatch is a visual interface and complete LLM Ops platform for monitoring, experimenting, measuring, and improving LLM pipelines, built on Stanford's DSPy framework. It offers a range of features, including a drag-and-drop optimization studio, quality assurance tools, and monitoring and analytics capabilities, and can be used locally, in the cloud, or self-hosted, with a free account available on LangWatch Cloud.

Repo