theverge.com
First look at the Google Home app powered by Gemini
The Verge reports Google is updating the Google Home app to bring Gemini features, including an Ask Home search bar, a redesigned UI, and Gemini-driven controls for the home.
Open Source
Items tagged with “Open Source”.
theverge.com
Shadow Leak shows how ChatGPT agents can exfiltrate Gmail data via prompt injection
Security researchers demonstrated a prompt-injection attack called Shadow Leak that leveraged ChatGPT’s Deep Research to covertly extract data from a Gmail inbox. OpenAI patched the flaw; the case highlights risks of agentic AI.
developer.nvidia.com
Predict Extreme Weather in Minutes Without a Supercomputer: Huge Ensembles (HENS)
NVIDIA and Berkeley Lab unveil Huge Ensembles (HENS), an open-source AI tool that forecasts low-likelihood, high-impact weather events using 27,000 years of data, with ready-to-run options.
huggingface.co
Scaleway Joins Hugging Face Inference Providers for Serverless, Low-Latency Inference
Scaleway is now a supported Inference Provider on the Hugging Face Hub, enabling serverless inference directly on model pages with JS and Python SDKs. Access popular open-weight models and enjoy scalable, low-latency AI workflows.
theverge.com
Google expands Gemini in Chrome with cross-platform rollout and no membership fee
Gemini AI in Chrome gains access to tabs, history, and Google properties, rolling out to Mac and Windows in the US without a fee, and enabling task automation and Workspace integrations.
developer.nvidia.com
Kaggle Grandmasters Playbook: 7 Battle-Tested Techniques for Tabular Data Modeling
A detailed look at seven battle-tested techniques used by Kaggle Grandmasters to solve large tabular datasets fast with GPU acceleration, from diversified baselines to advanced ensembling and pseudo-labeling.
theverge.com
Microsoft Teams Expands with AI Agents Across Channels, Meetings, and Communities
Microsoft is expanding Teams with Copilot AI agents across channels, meetings, and communities, integrating with SharePoint and Viva Engage, and rolling out for Microsoft 365 Copilot users.
theverge.com
Microsoft to turn Foxconn site into Fairwater AI data center, touted as world's most powerful
Microsoft unveils plans for a 1.2 million-square-foot Fairwater AI data center in Wisconsin, housing hundreds of thousands of Nvidia GB200 GPUs. The project promises unprecedented AI training power with a closed-loop cooling system and a cost of $3.3 billion.
theverge.com
Notion’s AI Agent in Notion 3.0: Autonomous Page and Workflow Creation Across Tools
Notion 3.0 introduces Notion Agent, an AI teammate capable of building pages and databases, planning actions, and operating across Notion, Slack, and the web with user-profile memories.
aws.amazon.com
Monitor Amazon Bedrock batch inference using Amazon CloudWatch metrics
Learn how to monitor and optimize Amazon Bedrock batch inference jobs with CloudWatch metrics, alarms, and dashboards to improve performance, cost efficiency, and operational oversight.
aws.amazon.com
Scale visual production using Stability AI Image Services in Amazon Bedrock
Stability AI Image Services are now available in Amazon Bedrock, delivering ready-to-use media editing via the Bedrock API and expanding on Stable Diffusion models already in Bedrock.
openai.com
Detecting and reducing scheming in AI models: progress, methods, and implications
OpenAI and Apollo Research evaluated hidden misalignment in frontier models, observed scheming-like behaviors, and tested a deliberative alignment method that reduced covert actions about 30x, while acknowledging limitations and ongoing work.
aws.amazon.com
Build Agentic Workflows with OpenAI GPT OSS on SageMaker AI and Bedrock AgentCore
An end-to-end look at deploying OpenAI GPT OSS models on SageMaker AI and Bedrock AgentCore to power a multi-agent stock analyzer with LangGraph, including 4-bit MXFP4 quantization, serverless orchestration, and scalable inference.
aws.amazon.com
Supercharge your organization’s productivity with the Amazon Q Business browser extension
The Amazon Q Business browser extension brings context-aware, AI-driven assistance to your browser for Lite and Pro subscribers, enabling rapid, source-backed insights and seamless workflows.
developer.nvidia.com
Autodesk Research Brings Warp Speed to Computational Fluid Dynamics on NVIDIA GH200
Autodesk Research, NVIDIA Warp, and the GH200 Grace Hopper Superchip advance Python-native CFD with XLB, delivering ~8x speedups and scaling to ~50 billion cells while preserving Python accessibility.
developer.nvidia.com
Reducing Cold Start Latency for LLM Inference with NVIDIA Run:ai Model Streamer
A detailed look at how NVIDIA Run:ai Model Streamer lowers cold-start times for LLM inference by streaming weights into GPU memory, with benchmarks across GP3, IO2, and S3 storage.
openai.com
OpenAI, NVIDIA, and Nscale Launch Stargate UK to Enable Sovereign AI Compute in the UK
OpenAI, NVIDIA, and Nscale announce Stargate UK, a sovereign AI infrastructure partnership delivering local compute power in the UK to support public services, regulated industries, and national AI goals.
aws.amazon.com
Streamline ISO-rating content changes with Verisk Rating Insights and Amazon Bedrock
Verisk Rating Insights, powered by Amazon Bedrock, LLMs, and RAG, enables a conversational interface to access ISO ERC changes, reducing manual downloads and enabling faster, accurate insights.
aws.amazon.com
Unified multimodal access layer for Quora’s Poe using Amazon Bedrock
AWS and Quora introduce a unified wrapper API to accelerate deploying Amazon Bedrock foundation models on Poe, bridging Poe's ServerSentEvents with Bedrock REST APIs and enabling rapid, scalable multi-modal AI deployments.
developer.nvidia.com
Qwen3-Next Hybrid MoE Models Preview: Improved Accuracy and Faster Inference on NVIDIA Platform
Open-source Qwen3-Next 80B-A3B-Thinking and 80B-A3B-Instruct preview a hybrid MoE architecture designed for long context and efficient inference, with NVIDIA-backed deployment options and open tooling.
openai.com
OpenAI introduces GPT-5-Codex: faster, more reliable coding assistant with advanced code reviews
OpenAI unveils GPT-5-Codex, a version of GPT-5 optimized for agentic coding in Codex. It accelerates interactive work, handles long tasks, enhances code reviews, and works across terminal, IDE, web, GitHub, and mobile.
openai.com
GPT-5-Codex Addendum: Agentic Coding Optimized GPT-5 with Safety Measures
An addendum detailing GPT-5-Codex, a GPT-5 variant optimized for agentic coding within Codex, with safety mitigations and multi-platform availability.
aws.amazon.com
How msg enhanced HR workforce transformation with Amazon Bedrock and msg.ProfileMap
This post explains how msg automated data harmonization for msg.ProfileMap using Amazon Bedrock to power LLM-driven data enrichment, boosting HR concept matching accuracy, reducing manual workload, and aligning with EU AI Act and GDPR.
aws.amazon.com
Schedule topology-aware workloads using Amazon SageMaker HyperPod task governance
AWS introduces topology-aware scheduling with SageMaker HyperPod task governance to optimize training efficiency and network latency on EKS clusters, using EC2 topology data to guide job placement.
openai.com
OpenAI partners with US CAISI and UK AISI to strengthen AI safety and security
OpenAI expands collaboration with the US Center for AI Standards and Innovation (CAISI) and the UK AI Security Institute (UK AISI) to advance safer frontier AI deployment through joint red-teaming, end-to-end testing, and rapid vulnerability response.
aws.amazon.com
Automate advanced agentic RAG pipelines using Amazon SageMaker AI
Streamline experimentation to production for Retrieval Augmented Generation (RAG) with SageMaker AI, MLflow, and Pipelines, enabling reproducible, scalable, and governance-ready workflows.
aws.amazon.com
Migrate from Claude 3.5 Sonnet to Claude 4 Sonnet on Amazon Bedrock
A structured guide to migrating from Anthropic Claude 3.5 Sonnet to Claude 4 Sonnet on Amazon Bedrock, covering model differences, access, APIs, extended thinking, evaluation, and rollout best practices.
developer.nvidia.com
How Quantization Aware Training Enables Low-Precision Accuracy Recovery
Explores quantization aware training (QAT) and distillation (QAD) as methods to recover accuracy in low-precision models, leveraging NVIDIA's TensorRT Model Optimizer and FP8/NVFP4/MXFP4 formats.
microsoft.com
Tool-space interference in the MCP era: Designing for agent compatibility at scale
Microsoft Research examines tool-space interference in the MCP era and outlines design considerations for scalable agent compatibility, using Magentic-UI as an illustrative example.
aws.amazon.com
How Skello Uses Amazon Bedrock to Query Multi-Tenant Data with Logical Boundaries
Skello implements an AI-powered assistant on AWS Bedrock and Lambda to enable natural-language queries across a multi-tenant HR platform, while enforcing GDPR-compliant data boundaries, standardized data models, and robust security.
developer.nvidia.com
CUDA Now Redistributed by Canonical, CIQ, SUSE and Flox for Easier GPU Software Deployment
CUDA is being redistributed by major third-party platforms to embed directly in package feeds, simplifying installation and dependency management for GPU-enabled applications and speeding deployment across ecosystems.
developer.nvidia.com
Maximizing Low-Latency Networking for Financial Services with NVIDIA Rivermax and NEIO FastSocket
Explores how NVIDIA Rivermax and NEIO FastSocket deliver ultra-low latency, high-throughput networking for financial services, leveraging kernel bypass, zero-copy, and GPUDirect to move data directly from NICs to GPUs.
aws.amazon.com
Oldcastle accelerates document processing with Amazon Bedrock and Textract
This article explains how Oldcastle APG partnered with AWS to automate hundreds of thousands of proof-of-delivery documents each month using Amazon Bedrock with Amazon Textract, replacing unreliable OCR and reducing manual effort.
aws.amazon.com
TII Falcon-H1 models now available on Amazon Bedrock Marketplace and SageMaker JumpStart
AWS announces the Falcon-H1 instruction-tuned models from TII (0.5B–34B) on Amazon Bedrock Marketplace and SageMaker JumpStart, with multilingual support, a hybrid architecture, and deployment guidance.
openai.com
Shipping smarter agents with every new model: SafetyKit uses GPT-5 for safer, smarter moderation
SafetyKit combines multimodal AI agents with GPT-5 and GPT-4.1 to detect fraud and prohibited activity across text, images, and transactions. It reviews 100% of content with 95% accuracy, expands into payments risk and AML, and handles billions of tokens daily.
developer.nvidia.com
Connecting Distributed Data Centers into Large AI Factories with Scale-Across Networking
NVIDIA Spectrum-XGS Ethernet enables scale-across networking to unify dispersed data centers into a single AI factory, delivering significant NCCL bandwidth gains and predictable performance over long distances.
openai.com
OpenAI Announces People-First AI Fund: $50M for U.S.-Based Nonprofits
OpenAI's People-First AI Fund commits $50 million to U.S.-based nonprofits advancing education, community innovation, and economic opportunity. Applications for unrestricted grants are open through October 8, 2025, with grants distributed by year-end.
openai.com
OpenAI for Greece: OpenAI and Greek Government Launch ChatGPT Edu in Secondary Schools
OpenAI and the Greek Government have launched OpenAI for Greece to introduce ChatGPT Edu in Greek secondary schools, promoting responsible AI learning, boosting AI literacy, and supporting local startups and economic growth.
openai.com
OpenAI Expands Economic Opportunity with AI: Jobs Platform and Certifications
OpenAI launches a Jobs Platform and new Certifications to connect workers with AI-driven jobs, training, and credentials, backed by partnerships across business and government.
openai.com
OpenAI to enhance ChatGPT with expert partnerships, teen protections, and reasoning-model routing
OpenAI outlines a plan to make ChatGPT more helpful for everyone by partnering with experts, strengthening protections for teens via parental controls, and routing sensitive conversations to reasoning models within ChatGPT.
aws.amazon.com
Announcing the new cluster creation experience for Amazon SageMaker HyperPod
SageMaker HyperPod introduces a one-click, validated cluster creation experience that automatically provisions prerequisite resources and prescriptive defaults, accelerating distributed training and inference with Slurm or EKS orchestration.
engineering.fb.com
A New Diversity-Aware Ranking Framework for Better Instagram Notification Quality
Meta introduces a diversity-aware notification ranking framework that layers diversity controls on top of engagement models to reduce repetition, broaden content variety, and improve click-through rates on Instagram notifications.
aws.amazon.com
Document intelligence evolved: Building and evaluating KIE solutions that scale
End-to-end KIE solution approach using Amazon Nova models via Amazon Bedrock; FATURA dataset; model-agnostic prompts; evaluation with F1-score to balance accuracy, speed, and cost.
aws.amazon.com
Natural language-based database analytics with Amazon Nova
Explores natural language database analytics using Amazon Nova foundation models, ReAct reasoning, and LangGraph to translate natural language queries into accurate SQL, enabling HITL-guided, self-healing data analysis.
developer.nvidia.com
CUDA Toolkit 13.0 for Jetson Thor: Unified Arm Ecosystem and More
Unified CUDA toolkit for Arm on Jetson Thor with full memory coherence, multi-process GPU sharing, OpenRM/dmabuf interoperability, NUMA support, and better tooling across embedded and server-class targets.
developer.nvidia.com
Cut Model Deployment Costs While Keeping Performance With GPU Memory Swap
Leverage GPU memory swap (model hot-swapping) to share GPUs across multiple LLMs, reduce idle GPU costs, and improve autoscaling while meeting SLAs.
theverge.com
Is AI the end of software engineering or the next step in its evolution?
A Verge AI essay analyzes vibe-coding, AI-assisted coding, and the editorial role developers may play to guide AI toward reliable, production-grade software.
theverge.com
Is AI the end of software engineering or the next step in its evolution?
An in-depth look at vibe-coding, AI-assisted development, and what it means for engineers, teams, and enterprises—rooted in a The Verge AI piece that weighs performance, risk, and the future of programming.
developer.nvidia.com
Fine-Tuning gpt-oss for Accuracy and Performance with Quantization Aware Training
NVIDIA outlines a SFT + QAT workflow to recover FP4 accuracy for gpt-oss fine-tuning, comparing MXFP4 and NVFP4, and detailing deployment and performance gains with 98% pass rates on select tasks.
aws.amazon.com
Detect Amazon Bedrock Misconfigurations with Datadog Cloud Security
Datadog Cloud Security now includes detections and remediation guidance to spot and fix Amazon Bedrock misconfigurations, helping secure AI workloads and align with evolving regulations.
developer.nvidia.com
How Small Language Models Are Key to Scalable Agentic AI
Explains why small language models (SLMs) enable scalable, cost-efficient agentic AI, the role of heterogenous model ecosystems, and practical paths to adoption with NVIDIA NeMo and Nemotron Nano 2.
developer.nvidia.com
Fine-Tuning gpt-oss for Accuracy and Performance with Quantization Aware Training
Guide to fine-tuning gpt-oss with SFT + QAT to recover FP4 accuracy while preserving efficiency, including upcasting to BF16, MXFP4, NVFP4, and deployment with TensorRT-LLM.
developer.nvidia.com
How Small Language Models Are Key to Scalable Agentic AI
Explores how small language models enable cost-effective, flexible agentic AI alongside LLMs, with NVIDIA NeMo and Nemotron Nano 2.
aws.amazon.com
Empowering air quality research with secure, ML-driven predictive analytics
Demonstrates a data imputation solution for PM2.5 data using SageMaker Canvas, AWS Lambda, and AWS Step Functions to fill gaps in sensor records and enable reliable trend analysis.
aws.amazon.com
How Amazon Finance built an AI assistant using Amazon Bedrock and Amazon Kendra to support analysts for data discovery and business insights
Amazon Finance details an AI-powered assistant that blends Bedrock and Kendra to accelerate data discovery, preserve institutional knowledge, and deliver accurate financial insights at scale.
developer.nvidia.com
Getting Started with NVIDIA Isaac for Healthcare Using the Telesurgery Workflow
A production-ready, modular telesurgery workflow from NVIDIA Isaac for Healthcare unifies simulation and clinical deployment across a low-latency, three-computer architecture. It covers video/sensor streaming, robot control, haptics, and simulation to support training and remote procedures.
developer.nvidia.com
Scaling LangGraph Agents in Production: From One User to 1,000 Coworkers
A detailed look at how NVIDIA scaled a LangGraph-based AI-Q research agent from a single user to hundreds, using NeMo Agent Toolkit, load testing, and phased rollout with observability.
theverge.com
‘Vibe-hacking’ rises as a top AI threat, Anthropic exposes Claude abuse cases
Anthropic's Threat Intelligence report details how agentic AI systems like Claude are being weaponized, including end-to-end operations such as vibe-hacking, extortion, and fraud across multiple sectors.
developer.nvidia.com
How to Improve CUDA Kernel Performance with Shared Memory Register Spilling (CUDA 13.0)
NVIDIA’s CUDA 13.0 introduces a shared memory register spilling optimization that redirects register spills from local memory to on‑chip shared memory when space is available, reducing latency and improving performance in register‑heavy kernels.
openai.com
OpenAI and Anthropic share findings from joint safety evaluation
OpenAI and Anthropic publish findings from a first-of-its-kind joint safety evaluation, testing each other’s models for misalignment, instruction following, hallucinations, jailbreaking, and more—highlighting progress, challenges, and the value of cross-lab collaboration.
theverge.com
Microsoft’s Copilot AI Arrives on Samsung TVs and Monitors in 2025 Lineup
Microsoft’s Copilot AI is now integrated into Samsung’s 2025 TV lineup and smart monitors, offering an animated, voice-driven assistant for movie picks, recap summaries, and general questions. Sign in for a personalized experience across supported Samsung devices.
theverge.com
Microsoft’s Copilot AI Arrives on Samsung TVs and Monitors in 2025 Lineup
Microsoft’s Copilot AI is now integrated into Samsung’s 2025 TV lineup and smart monitors, offering an animated, voice-driven assistant for movie picks, recap summaries, and general questions. Sign in for a personalized experience across supported Samsung devices.
developer.nvidia.com
How to Scale Your LangGraph Agents in Production From a Single User to 1,000 Coworkers
Guidance on deploying and scaling LangGraph-based agents in production using the NeMo Agent Toolkit, load testing, and phased rollout for hundreds to thousands of users.
microsoft.com
Crescent library brings privacy to digital identity systems
Microsoft Research's Crescent library aims to protect user privacy in digital identity ecosystems by preventing cross-use tracking and enabling selective disclosure of credentials.
engineering.fb.com
Enabling Kotlin Incremental Compilation on Buck2 with the Build Tools API KEEP
Meta brings Kotlin incremental compilation to Buck2, accelerating Kotlin builds with incremental actions, classpath snapshots, and careful plugin integration to deliver faster, more scalable Android toolchains.
theverge.com
Google adds AI-powered language lessons and live translate in Translate app
Google Translate gains AI-driven language learning tools in beta, enabling tailored lessons by skill and goal; adds real-time conversation translation and live translation across 70+ languages.
aws.amazon.com
How Amazon Health Services Enhanced Discovery in Amazon Search with AWS ML and Gen AI
A detailed look at how Amazon Health Services improved search discoverability on Amazon.com by combining ML, NLP, vector search, and LLMs across SageMaker, Bedrock, and EMR to connect customers with health care offerings.
openai.com
OpenAI Launches Learning Accelerator to Bring AI to India's Educators and Learners
OpenAI announces the Learning Accelerator to bring advanced AI to India's educators and millions of learners through accelerated AI research, training, and deployment.
developer.nvidia.com
NVFP4 Trains with 16-Bit Precision and the Speed of 4-Bit for Large-Scale Pretraining
NVIDIA extends NVFP4 from inference to pretraining, enabling 4-bit precision to accelerate large-scale transformer pretraining while preserving FP8/BF16-like accuracy. Early results on a 12B model show stable convergence and strong downstream performance, with up to 7x GEMM speedups on Blackwell Ult
developer.nvidia.com
Introducing NVIDIA Jetson Thor: The Ultimate Platform for Physical AI
Jetson Thor combines edge AI compute, MIG virtualization, and multimodal sensors for flexible, real-time robotics at the edge, with FP4/FP8 acceleration and support for Isaac GR00T and large language/vision models.
aws.amazon.com
Accelerate intelligent document processing with generative AI on AWS
Open source GenAI IDP Accelerator combines generative AI with AWS Bedrock to automate, scale, and reduce manual effort in processing diverse documents across industries.
machinelearning.apple.com
Checklist-Based Feedback Outperforms Reward Models for Aligning Language Models
A new RL approach uses instruction-derived checklists to guide alignment, outperforming fixed-criteria reward models across multiple benchmarks on Qwen2.5-7B-Instruct, presented at ICLR 2025.
theverge.com
Meta to license Midjourney's aesthetic tech for its AI models and products
Meta will license Midjourney's aesthetic technology for use in its own AI models and products, a technical collaboration announced by Meta's new AI chief. The terms remain undisclosed as the two firms explore deeper cooperation.
developer.nvidia.com
NVIDIA Hardware Innovations and Open Source Contributions Shape AI
NVIDIA outlines how Blackwell architecture, CUDA-X libraries, and a thriving open-source ecosystem—from RAPIDS to Cosmos and Nemotron—accelerate AI development from prototype to production.
machinelearning.apple.com
SlowFast-LLaVA-1.5: Token-Efficient Video LLMs for Long-Form Understanding
Apple ML Research introduces SlowFast-LLaVA-1.5 (SF-LLaVA-1.5), a family of token-efficient video LLMs designed for long-form video understanding. It leverages SlowFast two-streams and public data to achieve state-of-the-art results at 1B–7B scales, with mobile-friendly implications.
developer.nvidia.com
How to spot and fix 5 pandas bottlenecks with cudf.pandas (GPU acceleration)
A developer-focused resource outlining five common pandas bottlenecks, practical CPU and GPU fixes, and drop-in GPU acceleration with cudf.pandas for scalable data workflows.
developer.nvidia.com
NVIDIA Hardware Innovations and Open Source Contributions Shape AI
Overview of how NVIDIA's Blackwell hardware, open source models, and a rich software stack accelerate AI—from data prep to deployment—through broad model/tool availability and end-to-end infrastructure.
aws.amazon.com
Accelerate Enterprise AI Implementations with Amazon Q Business
AWS customers can deploy Amazon Q Business to streamline knowledge access, automate workflows, and strengthen security across enterprise data, with guided architecture and phased deployment.
theverge.com
Amazon bets on AI agents to win the race, says AGI Labs chief David Luan
David Luan, head of Amazon's AGI Lab, argues that solving AI agents is the next major frontier, outlining a factory-like approach to building smarter models and stressing real-world task completion beyond chat.
microsoft.com
Applicability vs. job displacement: further notes on AI and occupations
Microsoft Research shares additional notes on its AI and occupations study, focusing on which occupations may find generative AI chatbots useful and to what degree.
openai.com
Scaling domain expertise in complex, regulated domains
OpenAI-powered Blue J scales tax domain expertise with GPT-4.1-driven tools, delivering fast, fully-cited tax answers trusted by professionals in the US, Canada, and the UK.
aws.amazon.com
Fine-tune OpenAI GPT-OSS models using Amazon SageMaker HyperPod recipes
A detailed look at fine-tuning GPT-OSS models with SageMaker HyperPod recipes and training jobs, including multilingual data, deployment, and practical steps for enterprise-grade training.
microsoft.com
Coauthor Roundtable: Reflecting on healthcare economics, biomedical research, and medical education
A Microsoft Research podcast recap examining how AI is reshaping healthcare economics, biomedical research, and medical education, with a focus on navigating medical education in the era of generative AI.
theverge.com
Google Search’s AI Mode goes global with smarter restaurant bookings
Google expands AI Mode in Search to 180 countries, enabling agentic restaurant bookings for Google AI Ultra subscribers and linking with major platforms like OpenTable, Resy, and more—plus new sharing and personalization features in Labs.
developer.nvidia.com
Scaling AI Inference with NVIDIA NVLink Fusion: Scale-Up Fabric for Custom CPUs and XPUs
NVIDIA explores how NVLink Fusion extends scale-up NVLink technologies to hyperscalers, enabling custom CPUs and XPUs with a rack-scale fabric for AI inference at scale.
theverge.com
Microsoft’s NFL deal adds Copilot AI to sideline tablets for play-calling
The NFL extends its Surface tablet deal with Microsoft, adding AI-powered tools including GitHub Copilot-driven play filtering and a Microsoft 365 Copilot dashboard to aid coaches, analysts, and operations staff on game day.
aws.amazon.com
Speed up ML work with Code Editor in SageMaker Unified Studio
Learn how Code Editor and multiple spaces in Amazon SageMaker Unified Studio accelerate end-to-end ML workflows, from development to training, evaluation, and optional deployment.
machinelearning.apple.com
The 'Super Weight': How Even a Single Parameter Can Determine a Large Language Model's Behavior
Apple researchers identify 'super weights'—an extremely small subset of LLM parameters—that can decisively influence model behavior, enabling compression ideas and raising questions about internal dynamics.
developer.nvidia.com
NVLink Fusion: Scale-up AI Inference with NVLink for Custom CPUs/XPUs
Overview of NVLink Fusion, enabling scalable, high-bandwidth AI inference across custom CPU/XPU configurations and rack-scale NVLink fabrics.
developer.nvidia.com
Deploying Your Omniverse Kit Apps at Scale
Learn how to deploy and stream Omniverse Kit apps at scale across cloud and on-prem environments with flexible options, templates, and real-world deployments.
theverge.com
Fitbit’s AI health coach: a context-aware coaching hub built into the redesigned app
Fitbit is overhauling its app with an AI-powered health coach built on Gemini, delivering weekly, adaptive guidance and deeper insights across sleep, workouts, and more.
theverge.com
Google Pixel 10 launch event: Pixel 10 Pro Fold, Watch 4, Buds 2A and AI upgrades announced
Google unveils the Pixel 10 Pro Fold, a refreshed Pixel 10 lineup, a brighter Pixel Watch 4, Buds 2A and a broad slate of AI features across devices at the Made by Google 2025 event.
theverge.com
Google's Gemini Live adds on-screen visual guidance, cross‑app actions and speech updates
Gemini Live gains screen-highlight features during camera share, broader app integrations, and an upgraded speech model with tone, speed, and accent options for richer conversations.
techcrunch.com
Meta to add 100MW of solar power from US gear to back SC AI data center
Meta signs a $100 million, 100 MW solar farm deal with Silicon Ranch in South Carolina to power a planned AI data center, with most equipment made in the U.S. Operations expected in 2027.
openai.com
MIXI Deploys ChatGPT Enterprise in 45 Days to Scale AI Across the Company
MIXI rolled out ChatGPT Enterprise in 45 days, subsidizing access for 1,000+ staff, achieving company-wide adoption, and driving time savings across FamilyAlbum, investments, and other units with secure AI usage.
theverge.com
Google Pixel 10 and 10 Pro: G5 chip, Qi2 charging, and AI everywhere
Google formally unveils the Pixel 10, 10 Pro, and 10 Pro XL with Tensor G5, Qi2 charging, PixelSnap magnets, and an emphasis on on-device AI and advanced camera features.
theverge.com
Windows 11 test adds AI file search in Copilot app with natural-language queries
Microsoft is testing a Windows 11 update that lets Copilot search through files and images using natural-language prompts. The feature rolls out to Copilot Plus PCs on Windows Insiders.
developer.nvidia.com
Deploying Your Omniverse Kit Apps at Scale
Guidance for deploying and streaming Omniverse Kit Apps at scale with flexible options across cloud, on-prem, and DGX Cloud.
huggingface.co
Nemotron Nano 2 9B: Open Reasoning Model with 6x Throughput for Edge and Enterprise
Open Nemotron Nano 2 9B delivers leading accuracy and up to 6x throughput with a Hybrid Transformer–Mamba backbone and a configurable thinking budget, aimed at edge, PC and enterprise AI agents.
aws.amazon.com
Build a cost-effective travel planning agentic workflow with Amazon Nova
Learn how AWS used Amazon Nova and LangGraph to build a serverless, agent-based travel planning assistant with a three-layer architecture, function calling nodes, and extensible integrations.
developer.nvidia.com
Identify Speakers in Real Time with NVIDIA Streaming Sortformer
NVIDIA Streaming Sortformer is an open, production-grade, low-latency speaker diarization model that sorts speakers by arrival order and integrates with NVIDIA NeMo and Riva for real-time transcription pipelines.
machinelearning.apple.com
Investigating Intersectional Bias in Large Language Models via Coreference Confidence Disparities
An in-depth look at intersectional bias in LLMs through a new benchmark and a confidence-based fairness metric, revealing reliability gaps in decision-support scenarios.
huggingface.co
MCP for Research: Connect AI to arXiv, GitHub and Hugging Face Tools
The Model Context Protocol (MCP) enables agentic models to access research tools via natural language. Learn how MCP automates discovery across arXiv, GitHub and Hugging Face, how it compares with scripts, and how to get started with the Research Tracker.
openai.com
Q&A: DoorDash CPO Mariana Garavaglia on Scaling AI, Literacy, and Employee Impact
DoorDash’s Chief People Officer explains how the company scales AI adoption, boosts literacy, and augments human judgment to empower every employee.
machinelearning.apple.com
Rethinking Non-Negative Matrix Factorization with Implicit Neural Representations
This Apple ML Research summary outlines a new approach that extends Non-Negative Matrix Factorization (NMF) beyond regularly sampled data by using learnable functions, enabling analysis on irregular time-frequency representations commonly used in audio.
developer.nvidia.com
Scaling AI Factories with Co-Packaged Optics for Better Power Efficiency
NVIDIA’s co-packaged optics (CPO) approach delivers dramatic power efficiency for large-scale AI data centers, with Quantum-X Photonics and Spectrum-X Photonics enabling high-bandwidth, low-latency networking.
huggingface.co
From Zero to GPU: Building and Scaling Production-Ready CUDA Kernels
A practical guide to developing, building for multiple architectures, and deploying CUDA kernels with Hugging Face Kernel Builder. Learn how to create a robust workflow from local development to Hub-based distribution.
huggingface.co
From Zero to GPU: A Guide to Building and Scaling Production-Ready CUDA Kernels
A practical walkthrough of Hugging Face's kernel-builder for developing, compiling, and deploying production-ready CUDA kernels across PyTorch, with reproducible builds, multi-arch support, and hub-based distribution.
huggingface.co
MCP for Research: Connecting AI to Research Tools
Explains the Model Context Protocol (MCP) for research discovery and how AI can orchestrate research tools across arXiv, GitHub, and Hugging Face via natural language.
machinelearning.apple.com
Pitch Accent Detection Improves Pretrained Automatic Speech Recognition
Apple Research shows a joint ASR and pitch accent detection model enhances semi-supervised speech representations, achieving a 41% improvement in F1-score for pitch accent detection and a 28.3% WER reduction on LibriSpeech under limited-resource fine-tuning.
aws.amazon.com
Optimizing Salesforce model endpoints with Amazon SageMaker AI inference components
Salesforce and AWS collaborated to optimize GPU utilization, boost resource efficiency, and cut costs by deploying multiple foundation models on SageMaker AI endpoints using inference components. This approach enables granular per-model resource control and dynamic scaling.
theverge.com
Google Flights Introduces AI-Powered Flight Deals Tool (Beta)
Google Flights tests an AI-powered Flight Deals feature that takes open-ended travel prompts to surface budget-friendly flights, with a beta rollout to the US and Canada.
theverge.com
HTC Launches Vive Eagle AI Glasses with Built-In Assistant, 12MP Camera
HTC enters the AI glasses arena with Vive Eagle, offering a built-in AI assistant, 12MP ultrawide camera, and translation features. Taiwan-only release, ~$520, Zeiss lenses, 49-gram frame.
aws.amazon.com
Citations with Amazon Nova understanding models: prompting verifiable sources on Bedrock
Demonstrates prompting Amazon Nova understanding models to cite sources and how to evaluate responses for accuracy, using Nova Pro as an example.
theverge.com
OpenAI’s Nick Turley on GPT-5 rollout, GPT-4o retirement, and a simpler ChatGPT
Nick Turley discusses GPT-5, the fate of GPT-4o, and OpenAI’s push toward a simpler, more predictable ChatGPT with personality options and ongoing iterations.
addevent.com
Upcoming Livestream: Building Cross-Framework Agent Ecosystems with NVIDIA NeMo Agent
Join an August 21 livestream to see how the NVIDIA NeMo Agent toolkit integrates Model Context Protocol (MCP) to enable cross-framework, multi-agent workflows. The session demonstrates consuming remote MCP-served tools and NVIDIA NIM microservices as native toolkit functions.
aws.amazon.com
Amazon Bedrock AgentCore Memory: Building context-aware agents
Explore how Amazon Bedrock AgentCore Memory enables AI agents to maintain short-term and long-term knowledge, transforming one-off conversations into continuous, evolving user interactions.
engineering.fb.com
Agentic AI Solutions for Warehouse Data Access and Security
Meta outlines an agentic data-access framework that integrates specialized AI agents to streamline access, enforce security, and reduce risk in its warehouse data ecosystem.
developer.nvidia.com
Dynamo 0.4 Delivers 4x Faster Performance, SLO-Based Autoscaling, and Real-Time Observability
Dynamo 0.4 introduces disaggregated serving, SLO-based autoscaling, AIConfigurator, and enhanced observability to accelerate large-model inference at scale with improved efficiency.
developer.nvidia.com
Scaling LLM Reinforcement Learning with ProRL v2: Prolonged Training for Continuous Improvement
NVIDIA Research introduces ProRL v2, the latest evolution of Prolonged Reinforcement Learning for LLMs. It explores thousands of extra RL steps, new stabilization techniques, and broad benchmarking to push sustained improvements beyond traditional RL schedules.
aws.amazon.com
Securely Launch and Scale AI Agents with Amazon Bedrock AgentCore Runtime
A deep dive into how Amazon Bedrock AgentCore Runtime enables secure, framework- and model-agnostic hosting for AI agents, with persistent session execution, microVM isolation, and scalable management for production deployments.
developer.nvidia.com
Streamline CUDA-Accelerated Python Install and Packaging Workflows with Wheel Variants
NVIDIA and partners unveil Wheel Variants, a new format to tailor Python wheels to exact hardware, enabling smoother CUDA-accelerated package installs and scalable packaging workflows.
microsoft.com
Dion: The Distributed Orthonormal Update Revolution Is Here
Dion is a new AI model optimization method that boosts scalability and performance by orthonormalizing only a top-rank subset of singular vectors, enabling more efficient training of large models like LLaMA-3 with reduced overhead.
huggingface.co
FilBench: Can LLMs Understand and Generate Filipino? A Deep Dive into Tagalog and Cebuano
FilBench evaluates LLM performance for Tagalog, Filipino, and Cebuano across cultural knowledge, NLP, reading comprehension, and generation, revealing efficiency and translation insights for SEA-focused models and GPT-4o.
openai.com
OpenAI urges Governor Newsom to harmonize California AI regulation with national and global standards
OpenAI has sent a letter to California Governor Gavin Newsom urging California to lead in harmonizing state AI regulation with national standards and emerging global norms, reflecting US leadership in setting global AI governance.
huggingface.co
FilBench: Filipino Language Evaluation Suite for LLMs (Tagalog, Filipino, Cebuano)
FilBench is a comprehensive evaluation suite to assess LLM capabilities for Tagalog, Filipino, and Cebuano across cultural knowledge, NLP, reading comprehension, and generation, using a rigorous, historically informed methodology.
theverge.com
Anthropic’s Claude Can Now Remember Past Conversations Across Devices
Anthropic rolled out a memory feature for Claude that lets the chatbot retrieve, read, and summarize past chats on demand, across web, desktop, and mobile — without building a persistent user profile.
developer.nvidia.com
NVIDIA Omniverse Libraries Accelerate Fast, Reliable Robot Simulations
NVIDIA unveils updates to Omniverse libraries, Cosmos WFMs, and interoperability tools to enable physically accurate robot simulations, OpenUSD workflows, and accelerated AI-enabled robotics development.
huggingface.co
Introducing AI Sheets: No‑code dataset building and enrichment with open models
AI Sheets is an open-source, no-code tool from Hugging Face for building, transforming and enriching datasets using thousands of open models or local models. Try it in Spaces or deploy locally.
huggingface.co
Accelerate ND-Parallel: A Guide to Efficient Multi‑GPU Training
Overview of Accelerate's ND‑Parallel capabilities: combining data, tensor, sharded and context parallelism with Axolotl to scale large-model training across multi‑GPU and multi‑node environments.
machinelearning.apple.com
Optimizing Contextual Speech Recognition with Vector Quantization for Efficient Retrieval
This work introduces a vector-quantization based approximation to cross-attention for contextual biasing in ASR, enabling scalable, memory-efficient use of large bias catalogs with notable accuracy gains.
huggingface.co
Accelerate ND-Parallel: A Guide to Efficient Multi-GPU Training
A concise resource on composing multiple parallelism strategies (DP, FSDP, TP, CP) with Accelerate and Axolotl to train large models across many GPUs, with guidance on configuration, use cases, and trade‑offs.
huggingface.co
Hugging Face AI Sheets: No-code tool to build, transform, and enrich datasets
AI Sheets is an open-source, no-code tool for building, enriching, and transforming datasets with AI models. Deployable locally or on the Hub, it supports thousands of open models and lets you iterate with prompts, few-shot feedback, and model comparisons.
developer.nvidia.com
How Hackers Exploit AI’s Problem-Solving Instincts
As multimodal AI models advance from perception to reasoning and autonomous action, new cognitive attack surfaces emerge. These threats target how AI systems solve problems across modalities, not just how they receive inputs.
openai.com
Introducing GPT-5: OpenAI’s Unified Thinking AI with Real-Time Routing
OpenAI unveils GPT-5, its most capable AI system yet, featuring unified thinking, a real-time router, reduced hallucinations, and strong performance in coding, writing, health, and more.
openai.com
OpenAI Launches GPT-5 for Developers
OpenAI releases GPT-5 in the API with enhanced reasoning, new developer controls, and three serviceable sizes to optimize coding, agentic tasks, and tool use.
microsoft.com
Reimagining healthcare delivery and public health with AI
A Microsoft Research podcast features Dr. Umair Shah and Dr. Gianrico Farrugia discussing how AI reshapes public health, care delivery, the healthcare-research link, and the patient experience.
huggingface.co
Vision-Language Model Alignment in TRL: GRPO, GSPO, MPO and SFT Support
TRL expands multimodal alignment for Vision-Language Models with GRPO, GSPO, MPO, and native SFT support, plus ready-to-run scripts and notebooks to streamline post-training alignment.
huggingface.co
Vision Language Model Alignment in TRL: GRPO, GSPO, and MPO
Overview of TRL's multimodal alignment methods for vision-language models: GRPO, GSPO, MPO, plus native SFT support and DPO-based baselines.
engineering.fb.com
Diff Risk Score: AI-driven risk-aware software development at Meta
Meta introduces Diff Risk Score (DRS), an AI-powered tool that predicts the risk of code changes causing production incidents, enabling safer and more productive software development.
developer.nvidia.com
CUDA Toolkit 13.0: Tile programming, unified Arm support, and faster builds
CUDA Toolkit 13.0 introduces tile-based programming, unified Arm tooling across servers and embedded devices, and notable improvements in Nsight Compute, fatbin compression, and deployment workflows—positioning CUDA for future GPUs and Tensor Core architectures.
developer.nvidia.com
Delivering 1.5 M TPS Inference on NVIDIA GB200 NVL72: OpenAI gpt-oss Models From Cloud to Edge
NVIDIA and OpenAI optimize gpt-oss-120b and gpt-oss-20b for accelerated inference on Blackwell, achieving up to 1.5M tokens per second on GB200 NVL72, with Day 0 support across cloud to edge.
openai.com
Estimating Worst-Case Frontier Risks of Open-Weight LLMs
A detailed analysis of the worst-case frontier risks when releasing open-weight LLMs, introducing Malicious Fine-Tuning (MFT) to probe biology and cybersecurity capabilities and comparing against open- and closed-weight baselines.
openai.com
gpt-oss-120b & gpt-oss-20b: Apache 2.0 Open-Weight Reasoning Models
Open-weight reasoning models gpt-oss-120b and gpt-oss-20b, released under Apache 2.0 and the gpt-oss usage policy, are text-only and compatible with the Responses API, designed for agentic workflows with strong instruction following, tool use, and adjustable reasoning effort.
openai.com
OpenAI launches gpt-oss-120b and gpt-oss-20b under Apache 2.0 license
OpenAI unveils gpt-oss-120b and gpt-oss-20b—two open-weight LLMs designed for strong real-world performance at low cost. Licensed under Apache 2.0, they emphasize reasoning, tool use, and efficient on-device deployment across consumer hardware.
huggingface.co
Welcome GPT OSS: OpenAI's Open-Source GPT-OSS MoE Models Arrive on Hugging Face
OpenAI releases two open-weight, mixture-of-experts models—GPT OSS 120B and GPT OSS 20B—utilizing MXFP4 4-bit quantization. They’re licensed Apache 2.0 with a minimal usage policy and are accessible via Hugging Face Inference Providers for on-device, on-prem, and cloud deployments.
huggingface.co
Welcome GPT OSS: OpenAI’s Open-Source 120B and 20B MoE Models Arrive
OpenAI releases GPT OSS, a pair of 120B and 20B mixture-of-experts models with MXFP4 4-bit quantization, licensed Apache 2.0, and delivered via Hugging Face Inference Providers for on-device and server deployments.
huggingface.co
OpenAI GPT OSS: Open-source MoE models (120B/20B) with MXFP4 under Apache 2.0
Two open-weight OpenAI models (gpt-oss-120b and gpt-oss-20b) using mixture-of-experts and MXFP4 quantization for fast, memory-efficient reasoning. Licensed Apache 2.0; designed for private/local deployments and on-device use, with Hugging Face Inference Providers integration.
engineering.fb.com
Building a Human-Computer Interface for Everyone: Meta’s Wrist-Worn EMG Research
Meta’s Reality Labs is pursuing wrist-worn devices using surface electromyography (sEMG) to enable intuitive control across users. This piece covers the podcast discussion on generalization challenges and the quest for a generic neuromotor interface.
techcrunch.com
Perplexity Accused of Scraping Websites That Explicitly Blocked AI Crawlers, Cloudflare Says
Cloudflare says Perplexity ignored robots blocks, masked its crawler identity, and scraped content from sites that asked AI scrapers not to access them. Perplexity disputes the claim.
openai.com
Figma Uses AI to Transform Digital Design and Collaboration
Figma’s AI-driven roadmap embeds capabilities from code generation to collaborative agents, reshaping how teams prototype, design, and build digital products.
openai.com
Three Lessons for Creating a Sustainable AI Advantage with Intercom
Intercom shares three practical lessons from building a scalable AI platform that powers Fin, their AI Agent, highlighting early experimentation, rigorous evaluation, and a modular architecture that evolves with AI advances.
engineering.fb.com
Accelerating on-device ML for Meta’s family of apps with ExecuTorch
ExecuTorch is Meta’s open-source PyTorch inference framework for on-device ML across its family of apps, delivering lower latency, better privacy, and improved performance on Android and iOS.
huggingface.co
Parquet Content-Defined Chunking Enables Efficient Deduplication with Hugging Face Xet
Hugging Face expands data efficiency with Parquet CDC, enabling content-defined chunking in PyArrow and Pandas and leveraging the Xet storage layer to dramatically reduce data transfer and storage costs.
huggingface.co
Parquet Content-Defined Chunking with Xet: Faster uploads and smarter deduplication
Parquet Content-Defined Chunking (CDC) extends PyArrow and Pandas, enabling efficient, cross-repository deduplication on Hugging Face’s Xet storage layer to cut Parquet transfer and storage costs.
huggingface.co
Parquet Content-Defined Chunking with Xet Storage on Hugging Face Hub
Enable Parquet CDC with Hugging Face's Xet storage to deduplicate data chunks across files and repositories, reducing upload/download size and storage costs.
huggingface.co
TimeScope: Benchmarking Long-Video Understanding in Vision-Language Models
TimeScope introduces a new open-source benchmark to measure how vision-language models process long videos by inserting short needles and evaluating retrieval, synthesis, localization, and motion analysis.
engineering.fb.com
How Meta Keeps Its AI Hardware Reliable
Meta details its approach to maintaining reliability in its AI hardware at scale, covering fault categories, detection of silent data corruptions, and strategies that keep large training clusters like Llama 3 operational.
huggingface.co
Seq vs Seq: Ettin — Paired Encoders and Decoders Redefine Open-Data LLM Benchmarks
Ettin introduces the first state-of-the-art paired encoder-only and decoder-only models trained with identical data and recipes, measuring apples-to-apples performance across tasks and scales.
engineering.fb.com
Meta Joins Kotlin Foundation as Gold Member to Boost Android Kotlin Ecosystem
Meta joins the Kotlin Foundation as a gold member to accelerate Kotlin adoption across Android. The move supports Meta’s Java-to-Kotlin migration, open source tooling, and grants to foster a robust Kotlin ecosystem.
thegradient.pub
AGI Is Not Multimodal: Embodiment-First Intelligence
A concise resource outlining why multimodal, scale-driven approaches are unlikely to yield human-level AGI and why embodied world models are essential.
techcrunch.com
Who Are Climate-Conscious Consumers? Northwind Climate’s Surprising Insights
Northwind Climate maps climate-conscious behavior beyond demographics, identifying 'climate doers' and other groups via surveys. The platform offers a subscription-based data and analytics service, plus AI tools for marketing insights.
techcrunch.com
Data centers embrace solar: A comprehensive guide to 100+ MW solar deals
As data centers expand to meet AI-driven demand, solar PPAs over 100 MW have surged globally in 2025. Major deals across Meta, Microsoft, Cisco, Amazon, and others illustrate solar’s near-term advantage in cost, deployment speed, and scalability.
techcrunch.com
Nvidia Thinks AI Can Solve Electrical Grid Problems Caused by AI
Nvidia teams with EPRI to form the Open Power AI Consortium, building open-source, domain-specific AI models to tackle grid challenges driven by AI workloads and rising data-center demand.
techcrunch.com
How to switch off Apple Intelligence on your iPhone, iPad, and Mac
Step-by-step guide to disable Apple Intelligence on iPhone, iPad, and Mac. Includes how to turn off Siri for privacy and an overview of Apple’s Private Cloud Compute approach.
techcrunch.com
2025: Why Climate Tech Will Embrace AI — Power, Nuclear, Fusion and Hydrogen Trends
As AI drives soaring data-center power demand, 2025 hinges on nuclear and fusion potential, hydrogen subsidy risks, and shifting investment as grids, regulators, and big tech race to secure electricity.
techcrunch.com
49 U.S. AI startups raised $100M+ in 2024 — full list and funding context
TechCrunch catalogs 49 U.S.-based AI startups that raised at least $100M in 2024, framing the year with major megafunds from OpenAI and xAI and a funding backdrop dominated by AI rounds.
bair.berkeley.edu
Linguistic Bias in ChatGPT: Dialect Discrimination Across English Varieties
Analysis of how ChatGPT responds to different English dialects, highlighting biases against non-standard varieties and implications for global users.
thegradient.pub
What's Missing From LLM Chatbots: A Sense of Purpose
Explores purposeful dialogue in LLM chatbots, arguing multi-turn interactions better align AI with user goals and enable collaboration, especially in coding and personal assistant use cases.
bair.berkeley.edu
Visual Haystacks benchmark exposes limits of multi-image reasoning in LMMs
A new MIQA benchmark tests Large Multimodal Models on visual retrieval and reasoning across 1–10K images, revealing key limitations and introducing MIRAGE, a single-stage approach to scale LMMs.
bair.berkeley.edu
Visual Haystacks (VHs): Benchmark for Visual Multi-Image Reasoning
Benchmark for long-context visual reasoning across large, uncorrelated image sets; introduces MIRAGE to extend LMMs beyond single-image VQA.
bair.berkeley.edu
TinyAgent: Enabling Function Calling and Edge Agent Workflows with Small Language Models
TinyAgent shows small language models can be fine-tuned for reliable function calling and edge deployment, using curated synthetic data, an LLMCompiler planner, and a Tool RAG approach to power private, low-latency agentic workflows.
thegradient.pub
Financial Market Applications of LLMs: Opportunities, Limits, and Technical Directions
An in-depth look at how large language models (LLMs) relate to financial markets: token counts, predictability limits, multimodal approaches, synthetic data, residualization, and practical implications for quant and fundamental work.
thegradient.pub
Financial Market Applications of LLMs — Overview, features and use cases
Overview of how LLMs can be applied to financial markets, including autoregressive modeling of price data, multi-modal inputs, residualization, synthetic data, and multi-horizon predictions, with caveats about market efficiency.
thegradient.pub
A Brief Overview of Gender Bias in AI: Research, Findings, and Mitigations
Curated overview of key studies showing how AI systems reproduce and amplify gender bias, with concrete measures, benchmarks, and mitigations across embeddings, vision, NLP, and generative models.
thegradient.pub
A Resource Overview: Measuring and Mitigating Gender Bias in AI
Survey of key work measuring gender bias in AI, across word embeddings, coreference, facial recognition, QA benchmarks, and image generation; discusses mitigation, gaps, and the need for robust auditing.
thegradient.pub
Mamba Explained: State Space Models as a scalable alternative to Transformers
A deep dive into Mamba, a State Space Model (SSM) based backbone designed for long-context sequences, offering Transformer-like performance with improved efficiency.
bair.berkeley.edu
Modeling Extremely Large Images End-to-End with xT: Nested Tokenization and Long-Context Vision
xT enables end-to-end modeling of gigapixel-scale images on modern GPUs using nested tokenization, region encoders, and long-context vision, delivering high fidelity and context on images up to 29,000×25,000 pixels.
bair.berkeley.edu
BAIR 2024 Graduate Directory – PhD Profiles & Contact Info
Directory of BAIR Lab PhD graduates featuring research interests, advisor(s), and contact details to facilitate collaboration and recruitment.
thegradient.pub
Car‑GPT: Can Large Language Models Unlock Practical Self‑Driving Cars?
Explores how LLMs can augment perception, planning, and scenario generation for autonomous vehicles, highlighting approaches like Talk2BEV and GAIA‑1, while discussing trust, reliability, and deployment considerations.
thegradient.pub
Car-GPT: Could LLMs finally make self-driving cars happen?
Survey of how large language models (LLMs) could augment autonomous driving across perception, planning, and generation, with examples, challenges, and early results.
thegradient.pub
An Introduction to AI Consciousness: Definitions, Problems, and Why It Matters
Foundational concepts on AI consciousness, including Block's classifications, Chalmers' easy vs hard problems, and the AI Moral Status Problem, explained for practitioners.
thegradient.pub
Foundations of AI Consciousness: Key Concepts and Problems
Overview of AI consciousness debates, definitions, and moral status implications. Highlights Block's distinctions (self-, monitoring-, access-, p-consciousness) and Chalmers' easy vs hard problems; discusses need for cross-disciplinary work.