Notizie — Pagina128

Google has upgraded its NotebookLM mobile app with AI-driven flashcards, quizzes, and an enhanced chat experience. Previously limited to desktop, these study tools now let users create customized flashcards and quizzes from uploaded sources on the go. The update also improves the Gemini-powered conversation feature, providing a longer context window and better memory on mobile devices. Users can selectively choose which sources the AI draws from, giving greater control when working with multiple documents. The changes aim to make NotebookLM a more flexible, portable hub for processing notes, PDFs, transcripts, and other text‑based materials.Leggi di più

OpenAI's AI‑video app Sora has seen a surge in popularity on Android, with the first‑day download count far exceeding its iOS launch. The rapid adoption follows the removal of invite‑code restrictions and an expansion to more countries. Alongside the download spike, OpenAI is adding new creation tools such as expanded cameo options, basic video editing, and storyboarding for Pro users, while also updating its payment structure to offer on‑demand video generations. These moves signal OpenAI's push to broaden Sora's appeal beyond casual creators toward professional users.Leggi di più

The Laude Institute announced its inaugural Slingshots grant program, providing funding, compute power, and product support to 15 AI research projects focused on evaluation. The cohort includes initiatives such as the Terminal Bench coding benchmark, an updated ARC-AGI project, Formula Code from Caltech and UT Austin, and Columbia's BizBench. SWE‑Bench co‑founder John Boda Yang leads the new CodeClash competition framework. Recipients are expected to deliver tangible outcomes like startups or open‑source codebases, while the institute warns against benchmarks becoming overly company‑specific.Leggi di più

AI image‑generation tools have made impressive strides, but they continue to falter on several fronts. Reviewers note recurring problems with realistic human faces, trademarked logos, and dense compositions. While services such as Dall‑E 3, Midjourney, and Google’s Gemini‑powered Pixel tools can produce striking visuals, they often misrender expressions, miss brand details, or produce nonsensical overlapping elements. Users are advised to simplify prompts, adjust adjectives, and use post‑generation editing tools to correct errors. The ongoing challenges highlight both the rapid progress and the current limits of AI‑driven visual creation.Leggi di più

Google is expanding its Finance service with AI-driven capabilities that draw on crowdsourced prediction data from Kalshi and Polymarket. The new features let users ask finance‑related questions, such as future GDP growth, and receive answers that incorporate the latest market probabilities. The rollout begins in India and will reach U.S. users within the next few weeks, with early access available through Google Labs. Google emphasizes that the predictions are not guaranteed to be accurate, noting that only a small share of crypto wallets on Polymarket show profits.Leggi di più

OpenAI’s chief financial officer, Sarah Friar, sparked controversy after suggesting the company could benefit from a U.S. government backstop on its data‑center financing. She later walked back the remarks, stating OpenAI is not seeking such a guarantee. The clarification prompted reactions from OpenAI CEO Sam Altman, who reiterated the firm’s stance against taxpayer bailouts, and AI industry figure David Sacks, who emphasized that the government has no plans to rescue AI firms. The exchange highlighted the tension between massive infrastructure spending and financing strategies in the fast‑growing AI sector.Leggi di più

Meta announced the European rollout of Vibes, its short‑form video feed that showcases AI‑generated videos within the Meta AI app. The feature lets users create, remix, and share AI‑crafted clips, and cross‑post them to Instagram and Facebook Stories and Reels. The launch follows a U.S. debut six weeks earlier and arrives as OpenAI introduced its own AI video platform, Sora, while other social networks grapple with the surge of low‑quality AI content.Leggi di più

California startup Subtle Computing has developed a compact, low‑latency voice‑isolation model that dramatically improves speech recognition for AI applications in noisy settings. By tailoring models to specific device acoustics and preserving user voice characteristics, the company claims an order‑of‑magnitude performance gain over generic solutions. The technology runs on devices with just a few megabytes of memory and 100 ms latency, and it has earned a spot in Qualcomm’s voice and music extension program. Backed by $6 million in seed funding, Subtle Computing is positioning itself as a key enabler for consumer, automotive, and hardware partners seeking reliable voice AI experiences.Leggi di più

A comprehensive glossary presents sixty essential artificial‑intelligence terms, from foundational concepts like artificial intelligence and machine learning to specialized notions such as diffusion models, sycophancy, and paperclip maximizer. Each entry offers clear definitions, contextual examples, and links to related technologies, helping readers navigate the rapidly expanding AI landscape and understand the terminology shaping modern digital tools.Leggi di più

Apple has agreed to partner with Google to develop a custom Gemini AI model that will power the next generation of its Siri virtual assistant. The deal, valued at about $1 billion per year, will see the model run on Apple’s private cloud for complex tasks while Apple’s own models remain on devices for privacy‑focused functions. Apple considered Anthropic as an alternative but found Google’s offering cheaper. The arrangement reflects Apple’s strategy to accelerate AI capabilities without building a large model in‑house, while maintaining a close financial relationship with Google.Leggi di più

Google Finance is receiving a major AI upgrade with the introduction of a Gemini‑powered Deep Search tool that delivers fully cited, comprehensive answers and a research plan for users. The service will also incorporate prediction‑market data from Kalshi and Polymarket, letting traders query future market events and see crowd‑sourced probabilities. The Deep Search feature will roll out in the United States in the coming weeks, with early‑access options via Google Labs and higher usage limits for Google AI Pro and AI Ultra subscribers. An Indian rollout will launch later, offering English and Hindi support but without the new Deep Search capability.Leggi di più

Amazon is extending its next‑generation Alexa+ platform to BMW vehicles, allowing the automaker to enhance its Intelligent Personal Assistant with more natural, conversational capabilities. The integration builds on the existing Alexa Custom Assistant framework that BMW already uses, and will tap into a range of large language models to handle complex driver requests. The rollout will begin with select BMW models, though specific timing and model details have not been disclosed.Leggi di più

Microsoft AI chief Mustafa Suleyman revealed a new team dedicated to creating a "humanist superintelligence" that is deliberately limited in autonomy and designed to serve humanity. The initiative emphasizes keeping humans "at the top of the food chain" and rejects narratives of a race to artificial general intelligence. Suleyman outlined three primary applications: an AI companion for learning and productivity, support for the healthcare sector, and breakthroughs in clean‑energy science. The move follows a new agreement with OpenAI that lets Microsoft pursue AGI independently or with partners, while leveraging OpenAI’s intellectual property.Leggi di più

OpenAI’s board removed CEO Sam Altman in 2023 after concluding he was not consistently candid with them. A deposition by co‑founder Ilya Sutskever, part of Elon Musk’s lawsuit, detailed accusations that Altman pitted senior executives against each other, offered conflicting information, and pursued merger talks with Anthropic. Sutskever presented memos documenting these concerns, citing examples of alleged manipulation, a toxic workplace, and potential conflicts of interest. The fallout saw several top executives depart and the board undergo significant changes, underscoring a dramatic leadership crisis at the AI leader.Leggi di più

Taskade, founded in 2017, offers a cross‑platform workspace that blends tasks, notes, outlines, and multiple view options. In 2023 the company added AI features powered by OpenAI's GPT models, enabling project planning, automation triggers, and a "Project Studio" that can generate outlines, workflows, and even full‑stack apps from a single prompt. The platform includes over 300 templates, a chat and video interface for teams, and a new extension called Taskade Genesis that lets users create AI‑driven applications quickly. While the tool shines for early‑stage startups and remote teams, its breadth may exceed the needs of users seeking only basic task‑management.Leggi di più

Google has expanded Gemini’s Deep Research feature to tap into users' Gmail, Docs, Sheets, Slides, PDFs, and Chat history. The update lets the AI generate more personalized reports by combining public web sources with personal files, giving users the ability to request analyses that reflect their own data. Users can choose which Google services to connect, and the feature is currently available on desktop with mobile support slated for the near future.Leggi di più

Inception, an AI startup focused on diffusion‑based models, announced a $50 million seed round led by Menlo Ventures with participation from Mayfield, Innovation Endeavors, Microsoft’s M12 fund, Snowflake Ventures, Databricks Investment, and Nvidia’s NVentures. Angel investors Andrew Ng and Andrej Karpathy also contributed. The company, led by Stanford professor Stefano Ermon, aims to apply diffusion techniques—traditionally used for image generation—to software development and natural‑language tasks. Its Mercury model has already been integrated into several development tools, and the team claims diffusion models can deliver higher token‑per‑second throughput and lower latency than conventional autoregressive systems.Leggi di più

Apple is close to finalizing a multi‑year agreement with Google that would let the tech giant’s Gemini model power key functions of Siri. The deal, reported to be worth roughly $1 billion per year, is intended to address shortcomings in Apple’s own Apple Intelligence initiative, which has been criticized for limited capabilities. Internally, the project is dubbed “Glenwood,” while the new Siri version carries the codename “Linwood” and is slated for inclusion in an upcoming iOS release. Apple plans to blend Gemini’s capabilities with its private‑cloud infrastructure to preserve user privacy.Leggi di più

Google is rolling out its Gemini AI assistant within Google Maps for Android and iOS users in India, adding contextual navigation help, multilingual support, and new safety features such as accident‑prone stretch alerts. The update, built on extensive localization, also brings real‑time road‑closure data from the National Highways Authority of India and speed‑limit displays in several major cities.Leggi di più