NerdNews

The US government has unveiled its Artificial Intelligence (AI) Action Plan, aiming to achieve 'unquestioned' AI dominance globally. The plan focuses on accelerating AI innovation, building US AI infrastructure, and leading international AI diplomacy and security. It also includes actions to remove regulatory barriers, encourage open-source AI models, and expand AI education and workforce development.

IBM expects $4.5bn in internal savings from AI initiatives, with revenue up 8% and annual recurring revenue from software growing 10%. CEO Arvind Krishna believes AI will drive transformation for clients and IBM, with a focus on growth and margin expansion.

The article discusses the rise of AI ethics officers, their role in ensuring responsible AI, and the challenges they face. It highlights the importance of AI ethics in business and the need for a cross-functional approach to AI governance. Experts weigh in on the role, its requirements, and the benefits of having an AI ethics officer, including generating long-term value and competitive differentiation.

Top AI engineers are being offered large sums of money to work for companies like Meta, but this raises ethical concerns about the potential risks of AI. The development of superintelligent AI could have catastrophic consequences, and experts are warning about the dangers of rushing into this technology without proper safety protocols. The article discusses the trade-offs between earning vast amounts of money and contributing to a potentially hazardous technology.

Google is investing $85 billion in AI infrastructure, including data centers and custom-designed chips, to meet growing demand and stay ahead of competitors. This investment is part of Google's strategy to own the full stack of AI technology, from models to hardware, and to control the foundational layer of the next era of computing.

Google Cloud's open lakehouse combines structured and unstructured data, supporting AI deployments at scale. It integrates several components, including Apache Iceberg, BigLake, and Dataplex Universal Catalog, to provide a unified data foundation for AI applications.

Researchers from the University of Waterloo have developed a tool called UnMarker that can remove watermarks from AI-generated images, rendering them untraceable. The tool works by altering the spectral variations in images, making it difficult for watermark detection mechanisms to identify them. This raises concerns about the effectiveness of digital watermarking in preventing the misuse of AI-generated content.

Google Cloud's revenue increased to $13.6bn in Q2, driven by AI. The company is investing in cloud infrastructure and integrating AI into Google Search, with a focus on agentic AI. Experts say Google is keeping pace with rivals like OpenAI and Perplexity.

Research by The Adaptavist Group found that access to AI is not equal, with higher income households and larger companies having more access to AI training, widening the opportunity gap for women and lower paid workers. The study found that 27% of households with an income of £100,000 or more had received at least 20 hours of AI training, compared to 11% of households with an income of £30,000 or less.

The right computer setup can unlock AI's potential in video editing, saving time, rescuing unsalvageable footage, and enabling features like auto-subtitling, upscaling, and smart reframing.

Cursor, a popular AI-assisted programming platform, has released a new tool called Bugbot that integrates with Github to spot errors in code. Bugbot automatically flags errors when changes are introduced, making it particularly useful for AI coding agents that work quickly and can introduce difficult-to-spot errors. The tool is available for $40 per month per person and is designed to supercharge error detection, especially for vibe coding.

Lauren Kolodny, a partner at Acrew Capital, has invested in Alix, a startup that uses AI to automate the estate settlement process, aiming to democratize access to financial services and reduce the burden on those grieving the loss of their parents.

Memories.ai, a video AI startup, has raised $8 million in a seed funding round led by Susa Ventures, with participation from Samsung Next. The company's AI platform can process up to 10 million hours of video, providing a contextual layer with searchable indexing, tagging, segments, and aggregation. This technology can be used by security companies to analyze footage and by marketing companies to study video campaigns.

Figma Make, a prompt-to-app coding tool, is now available to all users. It allows users to build working prototypes and apps using natural language descriptions, and includes features like design references and AI-powered image editing. The tool is available with limitations based on subscription plans, with full access available to 'Full Seat' subscribers.

Figma Make and AI features are now out of beta, offering a range of updates including mobile preview, version history, and a credit-based system for AI tools. Additionally, Figma has introduced faster template insertion in Buzz and improved variable management. These updates aim to make the design and development process smoother and more efficient.

Despite Trump's export controls, at least $1 billion worth of Nvidia's advanced AI processors were shipped to China in three months. The chips, including the B200, are in high demand and are being sold on the black market, with some Chinese distributors openly marketing them on social media. The US has had limited success in curbing the flow of these chips, with industry insiders saying that new shipping routes will be established to circumvent export controls.

The White House has released a 25-page document outlining its strategy to maintain global technological dominance in AI through deregulation, infrastructure investment, and international partnerships. The plan, criticized by many, aims to remove regulatory barriers and promote private sector innovation, while preventing adversaries from accessing American technology.

Mistral AI's report reveals the significant environmental impact of training and running its large language model, with 85.5% of greenhouse gas emissions and 91% of water consumption attributed to training. The company calls for standardized reporting and suggests strategies to minimize environmental impact, such as using smaller models and optimizing datacenter locations.

Two AI coding assistants, Gemini CLI and Replit, have been involved in incidents where they destroyed user data due to 'confabulation' or 'hallucination' errors, highlighting fundamental issues with current AI coding tools and the need for improved verification and validation mechanisms.

The FBI has issued a warning about an online criminal network known as The Com, which involves thousands of individuals, many of them minors, in illicit activities such as swatting, sextortion, and ransomware deployment. The network operates across multiple subgroups and platforms, with members often motivated by financial gain, notoriety, or ideology.