Search the archive

NVIDIA TO JOIN DOW JONES INDUSTRIAL AVERAGE, REPLACING RIVAL CHIPMAKER INTEL

Nvidia is replacing Intel in the Dow Jones Industrial Average, showing the growing influence of AI in the semiconductor sector.

GOOGLE'S 'BIG SLEEP' AI PROJECT UNCOVERS REAL SOFTWARE VULNERABILITIES

Google's AI discovered a new, exploitable bug in SQLite, showcasing how LLMs can identify real-world software vulnerabilities.

AMAZON WILL NOW USE AI TO RECAP WHAT YOU'RE WATCHING

Amazon's X-Ray Recaps is an AI tool that uses generative techniques for personalized TV show summaries on Prime Video.

GOOGLE IS OPENING AN AI HUB IN OIL-RICH SAUDI ARABIA

Google is building an AI-focused data center in Saudi Arabia to enhance research in Arab language AI models and Saudi-specific applications.

OPENAI'S O1 MODEL LEAKED ON FRIDAY AND IT IS WILD

AI FOR STARTUPS

Microsoft and a16z advocate for collaboration between Big Tech and Little Tech to foster AI innovation and competition. They support open-source AI and have proposed policy ideas to support startups and equalize opportunities in the AI economy. Their joint focus is on creating a robust, competitive ecosystem that leverages AI to drive economic growth and innovation.

META'S FORMER HARDWARE LEAD FOR ORION IS JOINING OPENAI

Caitlin Kalinowski is joining OpenAI to lead its robotics and consumer hardware efforts. OpenAI is reviving its robotics team to integrate AI into physical devices and collaborate on new AI hardware projects. Kalinowski's role will focus on partnerships to advance OpenAI's multimodal AI in hardware.

MV PAINT: TEXTURE GENERATION

MV Paint synchronizes the problem of texture and UV generation for 3D assets. It achieves strong performance.

OS-ATLAS: A FOUNDATION ACTION MODEL FOR GENERALIST GUI AGENTS

OS-ATLAS is a strong open foundation model for taking actions on a computer. Substantial effort has been put into making it work across different GUIs by collecting millions of screenshots on a variety of devices. The model outperforms essentially all other open models and many closed models at the task of GUI navigation.

SHORT AND LONG VIDEO UNDERSTANDING

PPLLaVA is a new model that excels in understanding both short and long videos, addressing a key challenge in video-based AI.

TENCENT HUNYUAN3D-1.0: A UNIFIED FRAMEWORK FOR TEXT-TO-3D AND IMAGE-TO-3D GENERATION

Hunyuan3D-1.0 is a strong new generative 3D model with powerful multi-view synthesis capabilities. The output models still aren't production ready, but they are a great starting point for artists looking to generate assets.

ANDROID LAB

Benchmark for autonomous agents on the Android mobile operating system.

AI'S SOCIAL SKILLS TO HELP HUMANS

Constrained Human-AI Cooperation (CHAIC) is a challenge designed to assess how well AI agents can cooperate with humans facing physical limitations.

HOW THE NEW YORK TIMES IS USING GENERATIVE AI AS A REPORTING TOOL

New York Times reporters used AI tools, particularly LLMs, to transcribe and analyze over 400 hours of audio for an investigation. Automated transcription significantly sped up the process, with LLMs identifying key themes and topics with improved accuracy. Human reporters were used to ensure the accurate interpretation and contextual understanding of the data, showing the importance of human-AI collaboration.

WRITING AS A WAY OF THINKING

This article discusses the impact of AI on writing and thinking, questioning the notion that writing is the only form of thinking. AI, such as ChatGPT, can enhance thinking through conversation. AI tools can augment rather than eliminate various thought processes. AI will change writing by automating mundane aspects, allowing for more creative and thought-provoking content.

WAYMO EXPLORES USING GOOGLE'S GEMINI TO TRAIN ITS ROBOTAXIS

Waymo is leveraging Google's MLLM Gemini to develop an end-to-end multimodal model for autonomous driving called EMMA. The project aims to enhance decision-making in the company's robotaxis. EMMA will integrate sensor data for better trajectory prediction and object detection, indicating a shift towards using LLMs beyond common applications. Further research is required to refine and scale these models for autonomous driving tasks.

users. Preview your docs on Mintlify

MORE THAN A QUARTER OF NEW CODE AT GOOGLE IS GENERATED BY AI

Over a quarter of new code at Google is now generated by AI, showing the technology's critical role in the company's operations.

META IS USING MORE THAN 100,000 NVIDIA H100 AI GPUS TO TRAIN LLAMA-4

Unlike closed models from other companies, Llama-4's open-source approach allows for widespread collaboration, with some commercial restrictions.

WONDER DYNAMICS NOW LETS YOU GO STRAIGHT FROM MULTI-CAMERA VIDEO TO FULLY ANIMATED 3D SCENE

Wonder Dynamics launched a tool that automates converting videos into fully editable 3D scenes.

CLAUDE 3.5 HAIKU NOW AVAILABLE

Claude 3.5 Haiku is slightly worse than GPT-4o and doesn't have vision capabilities, but it is still very intelligent and cheap relative to other models of its quality.

7 AI NEWS THAT GOOGLE ANNOUNCED IN OCTOBER

This article provides a recap of seven AI updates from October, covering Google Maps' biggest AI update, tips for getting started with NotebookLM, and more ways people can ask questions, search for information, and get an AI Overview.

META'S AI ABUNDANCE

Meta is uniquely positioned to capitalize on generative AI, especially in digital advertising. The company's investments in AI, including its Llama models, will support innovative advertising strategies like generative ads and AI-driven chat agents, potentially increasing demand and revenue by leveraging machine learning to enhance ad targeting and efficiency. Meta's focus on integrating AI across its platforms underscores its commitment to maintaining a competitive edge in the rapidly evolving AI landscape.

MODELS FOR PII DETECTION

The GLINER models and dataset are synthetic datasets that can be used for synthetic data.

RANDOMIZED AUTOREGRESSIVE MODELING FOR IMAGE GENERATION

This study introduces Randomized AutoRegressive (RAR) modeling for image generation, which achieves state-of-the-art results on the ImageNet-256 benchmark with an impressive FID score of 1.48.

MODELS WITH BUILT-IN INTERPRETABILITY

"B-cosification" is a method that adapts to existing pre-trained models to offer highly interpretable explanations of their predictions.

thoughts. Write a review

HERTZ DEV: OPEN SOURCE SPEECH-TO-SPEECH

Extremely impressive open release that is permissively licensed. This was trained to generate human speech from a variety of input modalities. The code is high quality and has some interesting tidbits around the encoder and decoder architectures.

RADIO PROPAGATION MODELING

This project presents a novel Machine Learning-aided Ray Tracing approach for radio propagation modeling designed to overcome the high computational demands of traditional methods.

GRAPH TOKENIZATION

GQT (Graph Quantized Tokenizer) is a novel tokenizer for graph data in geometric deep learning.

THE AI SERVICES WAVE: LESSONS FROM PALANTIR IN THE NEW AGE OF AI

AI is reshaping service industries, with companies like Palantir leading by integrating AI with operations to enhance scalability and efficiency. Startups are leveraging AI to automate and improve traditionally complex processes, creating significant value and transforming business models. The focus is on developing AI-driven "tech-services" that combine software capabilities with human expertise for better results and increased market competitiveness.

SAPIEN RAISES $8.7M SEED LED BY GENERAL CATALYST

Sapien is developing AI-powered financial analysis tools that transform complex, error-prone tasks into quick insights, revolutionizing CFO roles. The platform integrates data from disparate sources to provide dynamic and contextually aware analysis, aiming to eliminate human errors in financial processes. It recently raised $8.7M in funding. Sapien plans to expand and enhance its AI capabilities to empower finance teams across various industries.

HOW I WRITE CODE USING CURSOR

Cursor, a VS Code fork, integrates LLM-powered features like tab completion and chat interfaces, streamlining coding tasks by automating boilerplate and repetitive changes. While tab completion is efficient and fast, there are potential pitfalls such as occasional incorrect suggestions. The tool encourages new workflow patterns, reducing reliance on libraries for boilerplate and aiding quicker iteration across unfamiliar languages or frameworks.

ELEVENLABS HAS HIRED THE TEAM BEHIND OMNIVORE, A READER APP

ElevenLabs has acquired the Omnivore team to enhance its AI-powered reading app, ElevenReader.

LINKEDIN LAUNCHES ITS FIRST AI AGENT TO TAKE ON THE ROLE OF JOB RECRUITERS

LinkedIn's "Hiring Assistant" is an AI tool designed to streamline recruitment tasks for recruiters.

ANTHROPIC'S CLAUDE AI CHATBOT NOW HAS A DESKTOP APP

Anthropic has released a desktop app for its AI chatbot, Claude, on Mac and Windows.

META IS MAKING A ROBOT HAND THAT CAN 'FEEL' TOUCH

Meta is partnering with GelSight and Wonik Robotics to commercialize advanced tactile sensors for AI research.

AMD'S NEW LANGUAGE MODELS

AMD used the OLMO codebase to train and release a language model trained on its accelerators.

OASIS BY ETCHED & DECART

Etched and Decart's Oasis is an AI model that generates open-world games frame-by-frame from user inputs, showcasing advanced physical understanding and real-time interaction. Oasis uses a Diffusion Transformer architecture optimized for the upcoming Sohu chip, enabling large-scale, low-latency 4K video models. The model, which integrates custom hardware for improved performance at scale, represents a shift towards interactive AI-generated content.

OPENAI WILL START USING AMD CHIPS AND COULD MAKE ITS OWN AI HARDWARE IN 2026

OpenAI is collaborating with Broadcom and TSMC to develop custom AI chips which are expected to begin production by 2026. Meanwhile, OpenAI is integrating AMD's MI300 chips into its Microsoft Azure infrastructure. Despite aligning with other tech giants on custom chip strategies, OpenAI remains behind Google, Microsoft, and Amazon in hardware development.

3D SCENE RECONSTRUCTION WITHOUT CAMERA POSE

NoPoSplat is a feed-forward model that can reconstruct 3D scenes from sparse, multi-view images without needing accurate camera poses.

IMPROVING LANGUAGE MODELS' REASONING

This study tackles a unique challenge in LLMs: how well they handle reasoning prompts that contain irrelevant or incorrect rationale snippets.

LONG-CONTEXT EVALUATION IN LANGUAGE MODELS

This study identifies a key flaw in using perplexity (PPL) to evaluate LLMs' long-context capabilities - PPL averages across all tokens and misses crucial ones needed for understanding extended inputs. To improve this, the authors introduce LongPPL, a metric that highlights these essential tokens for a clearer measure of long-context performance.

Learn about their 90-day pilot