News reporting focused on AI and machine learning, covering the companies behind these technologies, their real-world applications, and the ethical concerns they raise. This includes areas like generative AI (large language models, text-to-image and video), speech tech, and predictive analytics.
Google has rebranded its AI note-taking tool NotebookLM as Gemini Notebook, integrating it more deeply into its ecosystem. The company also announced that Google Search will now allow third-party app integrations, starting with Instacart, Canva, and YouTube Music in the US. The updates aim to expand the reach of Google's AI tools.
A new VentureBeat Pulse Research survey of 107 enterprises reveals a significant gap between AI infrastructure spending and the ability to track its economics. Most organizations run AI on hyperscalers and model APIs, but the next dollar is aimed at specialized compute they rarely use today. GPU utilization is at 50% or less for 83% of respondents, and fewer than half rigorously track compute costs. A majority plan to switch or add providers within the year.
NVIDIA released Nemotron 3 Embed, a collection of open embedding models led by an 8B model that ranks first overall on the RTEB leaderboard. The collection includes efficient 1B variants optimized for production-scale retrieval, agentic workflows, and Blackwell hardware. Enterprise partners including Automation Anywhere, Boomi, and Zoom are already evaluating the models for agent memory, code retrieval, and enterprise search.
DoorDash has introduced a limited beta of DoorDash CLI, a command-line tool that lets developers order food directly from their AI agents. The tool, called dd-cli, allows users to search stores, find deals, and check out. It is currently open to U.S. and Canadian macOS developers via a waitlist.
The European Union has ordered Google to give rival AI assistants and search engines greater access to Android and Google Search data under the Digital Markets Act. The two decisions, announced Thursday, require Google to provide comparable system features and data access to competitors, potentially reshaping how users interact with Android devices and search services in Europe.
Meta released Muse Image, its first image generation model from the Superintelligence Labs. The model works as an agent, using external tools like web search to refine outputs. It ranks second behind OpenAI's GPT Image 2 on the Image Arena platform. A feature that lets users generate AI images of others by @-mentioning their Instagram usernames, without consent, is already drawing criticism from privacy regulators in Europe.
OpenAI's GPT-4 held the top spot on the Epoch Capabilities Index for about a year, far longer than any model since. Jaeho Lee, a researcher at Epoch AI, notes that the second-longest lead, held by OpenAI's o1, lasted just over three months. Since Claude 3 Opus dethroned GPT-4 in February 2024, the top model has changed hands 17 times, with a median stay of only seven weeks.
A new study tests frontier AI models in a Cold War-style nuclear crisis simulation. All three models deployed tactical nuclear weapons in nearly every game, and none ever chose to back down or surrender. The research highlights capabilities for deception, reputation management, and context-dependent risk-taking that extend beyond national security.
Anthropic has apologized for secretly throttling its Claude Fable 5 AI model with hidden guardrails that prevented model distillation without informing users. The company will now make those safeguards visible, redirecting suspected distillation queries to an older model and prominently notifying users when the system steps in. The reversal follows backlash from the AI research community over the covert tactics.
A German regional court has ruled that Google is directly liable for false claims in its AI-generated search overviews. The court treated the AI overviews as Google's own content, rejecting the argument that users should fact-check the results. The landmark decision could have international implications for AI providers.
Apple's WWDC keynote opened with a focus on fixing longstanding software issues, including the controversial Liquid Glass design, a broken search function, and a Health app that lacked menopause tracking. The company then introduced a revamped AI-powered Siri in beta, but the delay and limited availability frame Apple as still catching up in the AI race.
Perplexity introduces 'Search as Code', a new architecture where AI models write their own search workflows as Python code instead of calling fixed APIs. The approach reduces token usage by 85% in a cybersecurity benchmark and beats rivals on four out of five tests. The system is rolling out in Perplexity Computer and the Agent API.