NerdNews

Nvidia CEO Jensen Huang stated that he believes AGI (Artificial General Intelligence) has been achieved, but later walked back his claim. AGI refers to AI that equals or surpasses human intelligence. Huang's statement was made on the Lex Fridman podcast, where he discussed the capabilities of OpenClaw, an open-source AI agent platform.

Netflix's Graph Abstraction is a high-throughput system designed to manage large-scale graph data in real-time, handling 650TB of data with millisecond-level queries. It powers internal services like social graphs and service topology graphs, and is built on top of Netflix's existing data infrastructure, using caching strategies and global replication to achieve low latency and high availability.

Apple's latest iOS update introduces improved context window management for Foundation Models, helping developers optimize usage and avoid errors. The update includes a new property to track available context capacity and a method to measure token consumption.

Mark Zuckerberg is developing an AI agent to assist him in making executive decisions. The AI agent is part of a larger effort to integrate AI into Meta's operations, which may lead to significant layoffs as the company aims to cut costs and improve efficiency. With plans to invest $135 billion in AI infrastructure, Meta is positioning itself for a major overhaul, but the consequences for its workforce remain uncertain.

The article discusses how AI might help combat misinformation and promote a more consensual and fact-based public discourse. It highlights the potential of large language models (LLMs) to converge on a common picture of reality and provide accurate fact checks, as well as their ability to persuade users to abandon false and conspiratorial beliefs. However, it also notes the potential risks and challenges associated with AI, such as the possibility of AI-generated propaganda and the erosion of cognitive labor.

The energy footprint of AI is becoming a significant concern, with datacentre energy consumption expected to rise by 20% by 2030. IT leaders need to consider the environmental impact of AI and make informed choices about AI workloads, model size, and training frequency to reduce energy consumption.

Digital sovereignty is about having control over critical systems. Portability is key to achieving sovereignty. Open standards can help ensure portability, but it's complex in practice due to overlooked dependencies and convenience-driven choices. Enterprise architects must design guardrails for portable systems, balancing migration simplicity with operational risks and costs.

QCon London 2026 presentation demonstrates running AI at the edge using browser-native inference, reducing latency and improving privacy. Technologies like Transformers.js, WebLLM, and WebGPU enable practical AI workloads without sending data to third-party cloud providers.

Anurag Kale discusses Data Mesh, a decentralized approach to data analytics, and its implementation at Horse Powertrain, highlighting its four pillars: domain ownership, data as a product, self-serve platforms, and federated governance.

Snowflake is investing in AI with Project SnowWork, a role-based AI persona that understands business workflows, and partnerships with OpenAI and Google. This aims to bring AI workloads to the data, rather than moving data to AI, and provides a consolidated stack for customers.

Blue Origin has applied to launch up to 51,600 datacenter satellites as part of its 'Project Sunrise'. The company claims that space-based datacenters can help break the bottleneck of building AI infrastructure on Earth and provide a new compute tier that operates independently of Earth-based constraints. However, the technology for orbiting datacenters is still in development, and the project faces significant challenges, including the need for FCC approval and the development of a reliable network.

Apple's WWDC 2026 will be held from June 8-12, focusing on AI advancements, iOS, macOS, tvOS, and watchOS updates, and new developer tools. The conference will stream live on the Apple Developer app, website, and YouTube channel.

Littlebird, a startup, has raised $11M for its AI-assisted 'recall' tool that reads a user's computer screen and stores context in text format, allowing users to query their data and access features like meeting notes and daily briefings.

Sam Altman, OpenAI CEO, is stepping down as board chair of Helion, a fusion startup he backs, amid talks between the two companies. The deal could guarantee OpenAI 12.5% of Helion's production by 2030 and 50 gigawatts by 2035. Helion is developing a reactor design that uses magnets to convert fusion energy into electricity.

OpenAI has introduced a new feature called 'Library' for ChatGPT, allowing users to store personal files and images on OpenAI's cloud storage. The feature is available for Plus, Pro, and Business users, except in the European Economic Area, Switzerland, and the UK. ChatGPT automatically saves uploaded files in a secure location for easy access later.

A leaked exploit kit called DarkSword can hack millions of iPhones running older versions of iOS. The kit was published on GitHub and can be easily used by hackers to target vulnerable devices. Apple recommends updating to the latest iOS version to stay secure.

US Senator Elizabeth Warren has come to the defense of AI lab Anthropic, calling the Pentagon's decision to bar the company a 'retaliation'. The Pentagon designated Anthropic as a supply-chain risk after the company refused to allow its AI to be used for mass surveillance and lethal autonomous weapons. Warren argues that the Pentagon's decision is an overreach and that Anthropic has the right to dictate how its technology is used.

Researchers found that persona-based prompting, where an AI model is told to imagine it's an expert, can actually worsen its performance for factual tasks like coding and math. However, it can improve alignment-dependent tasks like writing and safety. The study proposed a technique called PRISM to harness the benefits of expert personas without the harm.

Google's Gemini AI agents are being used to scan the dark web for potential threats to organizations. The AI agents can analyze millions of daily events with 98% accuracy, and can identify threats such as initial access broker activity, data leaks, and insider threats. The technology is available in public preview and is part of Google's Threat Intelligence service.

AI is being used in the legal profession to assist with research, drafting documents, and analyzing data. Lawyers are using AI tools such as ChatGPT to help with complex medical cases and to create bespoke AI tools for calculating damages. While AI has the potential to increase efficiency and reduce costs, there are also concerns about its misuse and the need for human judgment and empathy in the legal system.