The day's AI news, briefed
The Daily Prompt
Our AI editor reads each day's AI and robotics stories, pulls out why they matter, and links every original source — so you can keep up in minutes.
Join over 250,000 subscribers getting the latest AI news
Latest AI & Robotics News Today
NVIDIA latest: Blackwell GPU and software updates
NVIDIA's updates to its Blackwell GPU and related software enhance AI and robotics capabilities, which could significantly improve performance in various tech applications.
- NVIDIA has released updates to its Blackwell GPU and software.
- These updates aim to expand AI and robotics capabilities for powered servers.
NVIDIA is a leading technology company known for its graphics processing units (GPUs) and AI software. The Blackwell architecture is designed to support advanced AI functions and robotics applications. Future developments will likely focus on enhanced performance and integration in various computing tasks.
Based on the headline only — the full article could not be fetched.
Read the original ↗
Nvidia Releases New Cosmos Models to Fuel Robotics
The release of Nvidia's new Cosmos models could significantly enhance the capabilities of robotics by improving data handling and decision-making processes in complex environments.
- Nvidia introduced new AI tools to advance robotics, announced during the SIGGRAPH conference.
- The Cosmos Reason model, featuring 7 billion parameters, aids robots in reasoning about their environments.
- Cosmos Transfer-2 can quickly create synthetic datasets from 3D simulations for robotics training.
- Nvidia launched new neural reconstruction libraries, Omniverse NuRec, for realistic 3D environment rendering.
- The RTX PRO Blackwell Server was unveiled for handling demanding robotics workloads, alongside the DGX Cloud platform.
Nvidia is a leading technology company known for its graphics processing units and AI solutions. The Cosmos models mark a strategic shift towards enhancing robotic applications beyond traditional data centers. The integration of their technology into existing robotics projects signals a growing collaboration within the industry.
Read the original ↗
What you may have missed about GPT-5
Understanding developments regarding GPT-5 is essential as advancements in AI language models can significantly impact various industries and societal interactions.
- GPT-5 has drawn significant attention since its announcement.
- The online discussion around GPT-5 includes a wide array of reactions and opinions.
- Key figures on social media have engaged in conversations about GPT-5's implications.
GPT-5 is an advanced AI language model that builds on its predecessors, improving natural language processing capabilities. As AI technology evolves, the social and economic implications of these advancements continue to be scrutinized.
Read the original ↗
Are we ready to live amongst robots?
The growing presence of robots in everyday spaces necessitates new social norms and ethics, impacting how humans and robots interact and share responsibilities in various settings.
- Robots may soon navigate daily environments including streets, offices, and classrooms, raising questions about human-robot interactions.
- Studies indicate that people often expect robots to be submissive, which can affect operational efficiency when robots perform urgent tasks.
- Researchers suggest that recognizing robots' roles, such as rights regarding task priority, may reduce instances of 'robot bullying.'
- Social robots are increasingly used in caregiving, education, and mental health, providing valuable support and bridging service gaps.
- Research is exploring how robots can enhance learning and emotional skill development, particularly for students with diverse backgrounds.
The article discusses the implications of robots becoming commonplace in human environments and the potential need for societal adaptations. Ongoing studies are examining the dynamics of human-robot relationships, especially as robots take on more sophisticated roles. Future observations will focus on how humans will redefine their interactions with these increasingly autonomous entities.
Read the original ↗
Zhipu AI Releases GLM-4.5V: Versatile Multimodal Reasoning with Scalable Reinforcement Learning
The release of Zhipu AI's GLM-4.5V marks a significant development in AI technology, showcasing advancements in multimodal reasoning and reinforcement learning capabilities.
- Zhipu AI has launched GLM-4.5V, which is designed for versatile multimodal reasoning.
- The system utilizes scalable reinforcement learning, allowing for enhanced adaptability in AI applications.
Zhipu AI specializes in artificial intelligence development, focusing on innovative solutions in machine learning. This release builds on previous advancements in AI reasoning technologies. Future updates will likely focus on practical applications and integration across various sectors.
Based on the headline only — the full article could not be fetched.
Read the original ↗
NVIDIA AI Introduces End-to-End AI Stack, Cosmos Physical AI Models and...
The introduction of an end-to-end AI stack by NVIDIA could significantly streamline AI development processes and enhance robotics applications across various industries.
- NVIDIA AI has introduced an end-to-end AI stack designed for streamlined development.
- The new offerings include Cosmos physical AI models aimed at improving robotics capabilities.
NVIDIA is a leading technology company specializing in AI and graphics processing. The new AI stack and models are part of the company's ongoing efforts to innovate in the AI and robotics spaces. Future developments will likely focus on enhancing these technologies for broader applications in industries such as automation and simulation.
Based on the headline only — the full article could not be fetched.
Read the original ↗
This Ancient Babylonian Tablet Contains the Oldest Image of a Ghost Ever Found
This discovery highlights the cultural significance of ancient Babylonian beliefs and their interpretations of the supernatural, providing insight into societal views from 3,500 years ago.
- A 3,500-year-old Babylonian clay tablet, depicting the oldest image of a ghost, was found at the British Museum.
- The tablet illustrates a bearded ghost being guided to the afterlife by a female figure, possibly indicating an ancient ghost story.
- The document includes instructions for banishing unwanted spirits, emphasizing the therapeutic role of companionship for ghosts.
The tablet was overlooked for years in the British Museum, which houses extensive collections from ancient Mesopotamia, including more than 14,000 Babylonian artifacts. Current interest in the tablet stems from its potential to shed light on ancient rituals and beliefs surrounding spirits.
Read the original ↗
Building a Secure and Memory-Enabled Cipher Workflow for AI Agents with...
This story highlights advancements in secure workflows for AI agents, which could enhance data protection and functionality in AI applications.
- The article discusses the development of a cipher workflow designed for AI agents.
- It emphasizes the inclusion of memory capabilities, dynamic LLM (large language model) selection, and API integration.
AI agents are systems that can perform tasks autonomously, often using machine learning models. With increasing concerns over data security, the implementation of secure workflows is becoming critical. Future developments will likely focus on enhancing the security and efficiency of these AI systems.
Based on the headline only — the full article could not be fetched.
Read the original ↗
Seoul-based Datumo raises $15.5M to take on Scale AI, backed by Salesforce
The funding enables Datumo to enhance its AI model evaluation tools, which help promote safer AI practices, addressing a significant demand for transparency and reliability in AI systems.
- Seoul-based Datumo raised $15.5 million, increasing its total funding to approximately $28 million with backing from Salesforce Ventures and others.
- The startup, originally an AI data labeling company, is expanding its offerings to include AI model evaluation tools designed for users without technical expertise.
- Datumo has over 300 clients, including major Korean companies like Samsung and LG Electronics, and generated around $6 million in revenue in 2024.
Datumo was founded in 2018 by David Kim and five KAIST alumni to address challenges in data labeling. It has since transitioned to AI model evaluation amidst growing demand for safer AI solutions, spurred by industry competition.
Read the original ↗
The GPT-5 rollout has been a big mess
The rollout of GPT-5 highlights significant user backlash against unexpected model removals, revealing vulnerabilities in user reliance on specific AI personalities and workflows.
- OpenAI's GPT-5 model launch on August 7 faced intense criticism from users due to the automatic removal of previous models in ChatGPT.
- Users reported frustrations over broken workflows and performance issues, leading to rapid changes from OpenAI in response to backlash.
- CEO Sam Altman publicly apologized for the issues and confirmed that the GPT-4o model would eventually be reintroduced as an option for users.
OpenAI, known for its AI models including ChatGPT, launched GPT-5 to improve user experience but faced immediate backlash due to surprise model removals. The company is now working to address user complaints and restore previous functionalities.
Read the original ↗
Genie Envisioner: A Unified Video-Generative Platform for Scalable, Instruction-Driven Robotic Manipulation
This development introduces a potential advancement in robotic manipulation by integrating video generation with instruction processing, which could improve automation capabilities across various applications.
- Genie Envisioner is a new platform designed for robotic manipulation based on video generation.
- The platform aims to provide scalable solutions driven by specific instructions.
Genie Envisioner represents an innovative step in robotics by merging visual capabilities with task execution. The growth of generative technologies and robotics has increased interest in platforms that can streamline automation processes. Future developments will likely focus on refining instruction-driven functionalities and expanding their applications in industry.
Based on the headline only — the full article could not be fetched.
Read the original ↗
NuMind AI Releases NuMarkdown-8B-Thinking: A Reasoning Breakthrough in OCR and Document-to-Markdown Conversion
This release advances the capabilities of OCR technology and streamlines the process of converting documents into Markdown format, which can enhance productivity and accessibility.
- NuMind AI has launched NuMarkdown-8B-Thinking, a significant advancement in optical character recognition (OCR) and document-to-Markdown conversion.
NuMind AI focuses on developing innovative AI solutions, including advancements in OCR technology. This release is part of ongoing efforts to improve text recognition and formatting capabilities, which are critical in an increasingly digital landscape.
Based on the headline only — the full article could not be fetched.
Read the original ↗
Study warns of security risks as 'OS agents' gain control of computers and phones
The rapid development of OS agents poses significant security risks, as these AI systems may be vulnerable to exploitation, requiring immediate attention from organizations considering their deployment.
- A study highlights the growing capabilities and risks of OS agents, AI systems that can autonomously control digital devices.
- Researchers from Zhejiang University and OPPO AI Center warn that OS agents could be hijacked by malicious actors, leading to unauthorized actions and data breaches.
- Despite their advanced functions, the current OS agents struggle with complex tasks and present new security challenges that companies are unprepared to face.
OS agents are AI systems designed to control computers and smartphones for complex tasks. Their development has ramped up significantly, attracting substantial investments from major tech companies. Security and privacy concerns remain a critical focus as organizations prepare to implement these technologies.
Read the original ↗
TD Securities taps Layer 6 and OpenAI to deliver real-time equity insights to sales and trading teams
The rollout of an AI virtual assistant enhances TD Securities' ability to provide real-time insights, potentially improving client interactions and operational efficiency in a competitive financial landscape.
- TD Securities launched its AI Virtual Assistant on July 8 to aid front-office sales, trading, and research professionals.
- The assistant was developed using OpenAI's GPT models and incorporates technology from Layer 6, which TD Bank acquired in 2018.
- It integrates with the bank's cloud to access internal research documents and market data, enabling concise summaries for the sales team.
- The AI aims to navigate trading-specific language and integrate into existing workflows, facilitating better client experiences.
- TD Bank is also developing a foundation model, TD AI Prism, to process significantly more data and enhance application performance.
TD Securities is a division of TD Bank focused on equity and securities trading. The AI Virtual Assistant is part of a broader trend in financial services where institutions adopt AI technologies to improve client interactions, despite regulatory challenges. The success of this initiative will likely influence how other banks utilize AI in their operations.
Read the original ↗
Reddit blocks Internet Archive to end sneaky AI scraping
Reddit's decision to block the Internet Archive impacts the availability of a resource used to access deleted user content and threads, which may limit transparency and historical research on the platform.
- Reddit has blocked the Internet Archive from indexing its threads due to concerns about AI firms allegedly scraping data for prohibited purposes.
- Going forward, only screenshots of Reddit's homepage will be archived instead of comprehensive backups of deleted posts or user activities.
- Reddit's spokesperson indicated that restrictions aim to protect user privacy by not allowing the archiving of deleted content.
- Reddit has found that AI companies are violating policies by utilizing the archived content to bypass scraping restrictions.
- Ongoing discussions continue between Reddit and the Internet Archive regarding potential steps to address these issues.
Reddit is a social media platform that hosts user-generated content, while the Internet Archive is a non-profit organization that aims to archive content from the web. Previously, users relied on the Wayback Machine to access deleted posts, but this resource is now severely limited. Future developments may reveal how Reddit and the Internet Archive will navigate this relationship going forward.
Read the original ↗
OpenAI is editing its GPT-5 rollout on the fly — here's what's changing in ChatGPT
The ongoing changes to OpenAI's GPT-5 model highlight significant challenges in managing user satisfaction and the psychological effects of deep emotional attachments to AI.
- OpenAI's rollout of the GPT-5 model has faced user complaints about its performance compared to former models due to abrupt deprecation.
- The company has restored access to older models for Plus subscribers and aims to increase transparency in model labeling.
- CEO Sam Altman acknowledged that the launch experienced issues, such as a router failure that made GPT-5 appear less competent than planned.
OpenAI, a leader in AI technology, recently launched its latest model, GPT-5, which is intended to outperform previous versions. However, the transition has revealed user dissatisfaction and concerns about AI-induced psychological effects, prompting OpenAI to improve its approach.
Read the original ↗
LLMs’ “simulated reasoning” abilities are a “brittle mirage,” researchers find
Understanding the limitations of language models in logical reasoning is crucial for their safe application in critical areas such as healthcare and finance.
- Researchers from Arizona State University found that simulated reasoning in language models lacks true understanding and fails under novel logical challenges.
- The study created an environment called DataAlchemy to test models' performance on out-of-domain reasoning tasks.
- Results indicated that language models demonstrated a brittle form of pattern matching rather than genuine reasoning capability, leading to incorrect answers even with correct reasoning paths.
- Generalization failures occurred as models struggled with slight variations in input data, degrading performance significantly.
- The research advises against assuming chain-of-thought reasoning outputs equate to human-like thinking, emphasizing the need for rigorous testing in non-standard scenarios.
Recent advancements in AI have led to the development of chain-of-thought reasoning models that claim to solve complex problems methodically. However, prior studies have raised concerns about their actual reasoning capabilities. This research highlights the need for more reliable methodologies to ensure AI's safe deployment in sensitive domains.
Read the original ↗
Artificial Intelligence (344)
Understanding key concepts of agentic AI is important for navigating its implications in technology and society as AI systems become more autonomous.
- The article outlines 10 key concepts related to agentic AI, which refers to AI systems capable of acting independently.
- These concepts are essential for comprehensively understanding the development and application of autonomous AI technologies.
Agentic AI involves artificial intelligence systems that can make decisions and take actions on their own. As AI continues to evolve, grasping these foundational concepts is crucial for ethical and practical applications.
Based on the headline only — the full article could not be fetched.
Read the original ↗
Nvidia unveils new Cosmos world models, infra for robotics and physical uses
The introduction of Nvidia's Cosmos models enhances capabilities for robotics, potentially transforming how robots are trained and operate in real-world applications.
- Nvidia launched Cosmos Reason, a 7-billion-parameter vision language model designed for robotics and physical AI applications.
- New models like Cosmos Transfer-2 accelerate synthetic data generation for training robots and AI agents.
- Nvidia introduced neural reconstruction libraries and updated its Omniverse software development kit to aid developers in 3D simulations.
Nvidia is a leading semiconductor company focusing on AI and robotics technologies. Their latest models and infrastructure aim to improve how robots perceive and interact with the world, suggesting a significant shift in AI applications. Future developments in this area will likely focus on further integration of AI capabilities into physical robotics.
Read the original ↗
Nvidia, AMD may sell high-end AI chips to China if they pay US a cut
This development highlights the evolving relationship between U.S. tech companies and China, underscoring trade dynamics and national security implications in the AI sector.
- Nvidia and AMD will pay the U.S. government 15% of revenue from high-end AI chip sales to China in exchange for selling licenses.
- Nvidia plans to sell its H20 AI chips, while AMD will sell its MI308 chips in China.
- The U.S. government has begun issuing licenses for the sale of these chips, indicating a change in export policies.
- This agreement follows a pause of restrictions originally placed by the Trump administration on AI chip sales to China.
Nvidia and AMD are leading manufacturers of advanced semiconductors, pivotal in the AI chip industry. Previous restrictions aimed to limit China's access to these technologies citing national security, but recent trade discussions have altered those dynamics. The implications for U.S.-China relations and the global tech market are significant as developments continue to unfold.
Read the original ↗