NerdNews

Nvidia has struck a non-exclusive licensing agreement with AI chip competitor Groq, hiring its CEO and other employees. The deal is reportedly worth $20 billion, making it Nvidia's largest ever. Groq's LPU chip can run LLMs 10 times faster and using one-tenth the energy.

Salesforce has integrated ChatGPT to prevent customers from building their own integrations, which could lead to data leaks. The integration allows users to update CRM data directly from the bot and is aimed at reducing the risk of exposing Salesforce data to external AI models.

ServiceNow is acquiring OT security specialist Armis for $7.8 billion, extending its security, risk, and OT portfolios. The deal will enhance ServiceNow's AI Control Tower platform, providing insight into exposure and automated remediation at scale to reduce risk.

Pinterest users are frustrated with the increasing amount of low-quality, AI-generated content on the platform. This 'AI slop' is flooding users' feeds, making it difficult to find authentic and useful content. The issue is not only affecting user experience but also undermining the platform's original purpose as a marketplace for trading real-world inspiration.

OpenAI is testing a new ChatGPT feature called 'Skills', similar to Claude's feature. This will allow ChatGPT to learn specific abilities, workflows, and domain-specific knowledge. The Skills feature is codenamed 'hazelnuts' and will be available as slash commands, including a Skills editor and an option to convert a custom GPT into a skill.

The rise of AI agents, which can take actions and complete tasks on behalf of users, poses significant privacy risks as they require access to personal data and systems. Experts warn that these agents could lead to a new host of problems, including data leaks, misuse, and interception, and threaten existing security practices.

New York Times investigative reporter John Carreyrou and five other writers have filed a lawsuit against xAI, Anthropic, Google, OpenAI, Meta, and Perplexity, alleging copyright infringement over the use of their books to train large language models without permission.

Despite significant investments in AI, many companies are failing to see returns due to poor implementation and management, rather than technological limitations. Successful AI deployments require a focus on capability, not just technology, and often involve partnering with external experts and adapting workflows to integrate AI effectively.

AI coding agents use large language models to write code, run tests, and fix bugs. They work by interpreting tasks, assigning them to parallel models, and evaluating outputs. Techniques like context compression and multi-agent architectures help them manage complex tasks.

NIST and MITRE are partnering to create two AI centers focused on manufacturing and cybersecurity, with a $20m investment to develop technology evaluations and advancements to protect US dominance in AI innovation.

Waymo's self-driving vehicles got stuck at intersections during a blackout in San Francisco. The company explained that its vehicles treat dead stop lights as four-way stops, but a 'confirmation check' system caused congestion. Waymo is shipping a software update to help its robotaxis navigate disabled traffic lights more decisively.

Italy's Competition Authority has ordered Meta to suspend its policy banning rival AI chatbots from WhatsApp, citing potential abuse of market dominance and harm to competition. Meta has argued that its API isn't designed for chatbot distribution and plans to appeal.

Global AI spending is forecast to reach $1.5 trillion in 2025, but many are questioning the value of this investment as it has yet to show significant returns. Companies are starting to delay or reduce AI spending due to the lack of tangible benefits.

Garmin's Emergency Autoland system has been used in a real emergency situation for the first time, landing a small plane safely at Rocky Mountain Metropolitan Airport in Colorado. The system took control of the plane after the pilots lost communication with air traffic control due to a pressurization issue. The Autoland system decided on the most optimal place to land the plane based on factors such as runway length, distance, and fuel range.

Waymo is rolling out updates to its US fleet to counter future disruption caused by power outages. The updates will improve the cars' decisiveness when navigating dark intersections by sending additional context around regional outages. This comes after a power outage in San Francisco caused chaos, with Waymo's self-driving cars becoming unresponsive due to traffic signal failures.

Airbus is preparing to migrate its mission-critical workloads to a digitally sovereign European cloud, driven by the need for access to new software and concerns over data sovereignty. The company estimates an 80% chance of finding a suitable provider, with a contract worth over €50 million to be awarded in the next few months.

AWS has announced support for Nutanix's AHV hypervisor in its Storage Gateway, allowing users to connect their hypervisors to AWS's Simple Storage Service (S3). This move is seen as a strategic play by AWS to drive demand for its services and tap into the growing market of VMware alternatives. Nutanix, a key challenger to VMware, has been expanding its storage options and this integration is expected to boost its growth.

The US has shut down a platform used by cybercriminals to steal Americans' bank account credentials, resulting in estimated losses of $14.6 million. The platform, web3adspanels.org, was used to store and manipulate stolen credentials, which were then used to attempt to access bank accounts and authorize illegal transfers.

The SEC has charged several crypto firms for allegedly defrauding US retail investors of over $14m through an investment scam. The scheme used social media and messaging apps to promise AI-powered investment returns, but instead diverted funds through bank accounts and crypto wallets.

Microsoft announced a new feature to block external users from sending messages, calls, or meeting invitations to organization members via Teams, integrating with Defender for Office 365 and enabling admins to manage blocked contacts through the Microsoft Defender portal.