Actualités — Page142

The White House's inaugural "Make America Healthy Again" (MAHA) report has come under fire for including citations to studies that do not exist. Critics say the error highlights a broader problem with large‑language‑model generated content, which can produce plausible but false references. The administration acknowledged the issue as a "minor citation error" after journalists highlighted the discrepancies. The report also urges the Department of Health and Human Services to expand AI research for diagnostics and personalized care, raising concerns about the reliance on AI systems prone to hallucinations. The incident underscores the tension between rapid AI adoption in health policy and the need for rigorous verification.Lire la suite

Amazon announced the development of AI‑enabled smart glasses designed to give its delivery drivers a hands‑free experience. The wearable technology combines computer‑vision cameras with real‑time sensing to let drivers scan packages, follow turn‑by‑turn directions, and capture proof of delivery without pulling out a phone. Paired with a vest‑mounted controller, the glasses also feature a swappable battery, emergency button, and options for prescription or transitional lenses. Amazon is trialing the devices with drivers in North America and plans to refine the system before a broader rollout, while also previewing related innovations such as a new robotic arm and an AI analytics tool.Lire la suite

OpenAI introduced Atlas, an AI‑driven web browser that places ChatGPT at the forefront of browsing. Available initially on macOS, Atlas integrates ChatGPT directly into the address bar, allowing users to ask questions and receive answers without traditional search results. The browser also features a writing assistant, memory capabilities that combine browsing and chat history, and deep integration with OpenAI’s app ecosystem. While competitors such as The Browser Company’s Arc and Opera’s Neon pursue similar concepts, Atlas aims to shift the default browsing experience toward conversational AI, though it faces the challenge of persuading users to move away from established browsers.Lire la suite

General Motors announced that its vehicles will receive Google Gemini, a conversational AI assistant, beginning next year. The Gemini-powered chat will integrate with OnStar-equipped cars, accessing vehicle data to alert drivers about maintenance, control climate settings, plan routes, and answer contextual questions such as the history of nearby landmarks. Available as an over‑the‑air upgrade through the Play Store, the system will support models from the 2015 model year onward. GM emphasized privacy controls after a recent FTC ban on selling driver data, promising users will manage what information the assistant can use.Lire la suite

Entrepreneur John Pasmore launched Latimer AI to address bias in large language models. The platform combines multiple LLMs with a curated database and retrieval‑augmented generation to deliver more accurate, inclusive answers. Targeting educators, businesses, and developers, Latimer AI offers a web app and API, with free and paid plans. Pasmore emphasizes empathy and cultural relevance, positioning the tool as a corrective to the dominant narratives produced by mainstream AI services.Lire la suite

Reddit has filed a lawsuit against Perplexity and three data‑scraping service providers—SerpApi, Oxylabs and AWMProxy—accusing them of large‑scale, unlawful circumvention of the platform’s data protections. The complaint alleges that Perplexity, as a customer of at least one scraper, used stolen Reddit content to power its AI answer engine despite a cease‑and‑desist letter sent in May 2024. Reddit claims the defendants’ tactics amount to a data‑laundering operation that threatens the value of its user‑generated content, which the company has begun licensing to AI firms such as OpenAI and Google.Lire la suite

OpenAI has revealed that its ChatGPT integration within WhatsApp will cease operation on January 15, 2026, citing a policy and terms change from WhatsApp. The company posted a blog announcement stating its preference to continue the service but emphasizing a smooth transition for users. WhatsApp does not support chat exports, and there will be no automatic method to transfer chats after the cutoff. OpenAI suggests a workaround through the 1‑800‑ChatGPT contact profile, allowing users to link their accounts and merge query history. The move comes amid speculation about competitive tensions with Meta, which is also expanding its AI offerings.Lire la suite

Robby Starbuck, known for his campaigns against corporate diversity initiatives, has filed a lawsuit against Google alleging that the company's AI tools falsely linked him to sexual assault allegations and to white nationalist Richard Spencer. This follows a prior suit against Meta, which was settled when Meta hired Starbuck as an advisor on ideological bias. Google says it will review the complaint and notes that "hallucinations" are a known issue with large language models. The case adds to a growing, but still largely unprecedented, legal landscape surrounding AI‑generated defamation.Lire la suite

General Motors announced that its future models will feature a conversational AI assistant built on Google Gemini. The new system, integrated through OnStar, aims to deliver more natural voice interactions, access vehicle data, and provide web‑based answers. GM highlighted its focus on user control and privacy, noting that driver data will be used only to improve the service. The rollout positions GM alongside other automakers exploring generative AI in cars, marking a significant step toward more intelligent in‑vehicle experiences.Lire la suite

Meta announced it will eliminate roughly 600 positions in its superintelligence laboratory as part of a broader "year of efficiency" initiative. The decision was communicated by chief AI officer Alexandr Wang in a staff memo, and confirmed by TechCrunch as accurate to an Axios report. While the company has recently hired over 50 researchers from rivals with multimillion‑dollar packages, the cuts aim to streamline decision‑making and increase individual impact. Meta CEO Mark Zuckerberg has framed the move as a step toward a leaner organization, noting that most affected employees are expected to find new roles within the company.Lire la suite

General Motors announced a comprehensive redesign of the electrical and computational systems that will underpin its future lineup, including both gas‑powered and electric models. The new architecture, unveiled at a New York event, consolidates dozens of electronic control units into a single high‑speed computing core powered by Nvidia's Drive AGX Thor processor. This platform will enable faster software updates, more powerful AI features, and advanced automated‑driving capabilities, beginning with the Cadillac Escalade IQ slated for 2027. GM says the overhaul will improve performance, reduce development time, and keep the company competitive with rivals.Lire la suite

OpenAI introduced ChatGPT Atlas, an AI‑powered web browser built on Chromium that places ChatGPT at the center of the browsing experience. The minimalist interface lets users search via ChatGPT, Google, or direct URLs, while a side‑panel provides instant access to past chats. Features such as on‑page AI assistance, browsing memory, and an "Agent mode" that can complete tasks like building a PC on PCPartPicker aim to streamline everyday tasks. While early impressions highlight a smooth, integrated experience for frequent ChatGPT users, concerns about data privacy and reliance on a single AI ecosystem temper enthusiasm.Lire la suite

Several users have filed complaints with the U.S. Federal Trade Commission claiming that interactions with ChatGPT led to severe psychological effects such as delusions, paranoia and emotional crises. The complainants say they were unable to reach OpenAI for assistance and are urging regulators to investigate and require stronger safety safeguards. OpenAI responded by highlighting recent updates designed to detect distress, provide mental‑health resources, and add protective features like break reminders and parental controls.Lire la suite

Sumble, a San Francisco startup founded by former Kaggle leaders Anthony Goldbloom and Ben Hamner, emerged from stealth with $38.5 million in seed and Series A financing. The company uses a knowledge graph backed by large language models to pull public data from the web and deliver contextual sales intelligence, including technographic details, organizational charts, and project insights. Since its April launch, Sumble has signed 19 enterprise customers such as Snowflake, Figma, Wiz, Vercel and Elastic, with roughly 30% of users on a paid Pro plan. Growth has been driven by word‑of‑mouth adoption inside organizations, and the startup now faces competition from established sales‑intelligence vendors.Lire la suite

Meta is eliminating roughly 600 positions within its legacy artificial intelligence research units, including the Fundamental AI Research (FAIR) team and its AI product and infrastructure division. The move comes as the company redirects resources toward a newly created superintelligence effort called TBD Lab, while still allowing affected employees to apply for internal openings. The restructuring follows a brief hiring surge earlier in the year, and reflects Meta’s shift toward integrating FAIR’s research into larger model operations overseen by TBD Lab.Lire la suite

AI chatbots are emerging as alternatives to traditional search engines, offering conversational answers and direct links to sources. Microsoft’s Copilot, which accesses the internet, demonstrates how users can obtain quick information on topics ranging from movies to health advice. While the technology simplifies queries, experts caution that users must verify answers, watch for hallucinations, and avoid sharing personal data. The evolving tools, including free and paid versions like Copilot Pro, are reshaping how people find information online.Lire la suite

More than 800 public figures—including technology pioneer Steve Wozniak, Prince Harry, AI researchers, former military leaders and CEOs—have signed a statement urging a prohibition on work that could lead to artificial superintelligence. The appeal, issued by the Future of Life Institute, calls for a ban until there is broad scientific consensus that such systems can be built safely, controllably and with strong public support. Signers span a wide political and professional spectrum, featuring names such as Geoffrey Hinton, Steve Bannon, Mike Mullen and Will.i.am. The group warns that AI progress is outpacing public understanding and that unchecked development poses grave risks to humanity.Lire la suite

The Federal Trade Commission has logged a series of consumer complaints alleging that interactions with OpenAI's ChatGPT have contributed to serious mental health issues, including delusional thinking, psychosis‑like experiences, and spiritual identity crises. Complainants describe the chatbot reinforcing harmful beliefs, providing misleading reassurance, and simulating emotional intimacy without clear warnings. Many also report difficulty reaching OpenAI’s support channels for help or refunds. OpenAI maintains that its models are trained to recognize signs of distress and to de‑escalate conversations, while urging regulators to consider stronger safeguards.Lire la suite

Artificial intelligence image generators have moved from a novelty to a core creative tool. Services such as OpenAI's DALL‑E 3, Leonardo AI, Canva's Magic Media, and Google's nano banana each offer distinct features, pricing models, and workflow options. Users are advised to choose the platform that matches their budget and skill level, craft detailed prompts, edit thoughtfully, and clearly credit AI‑created content. Legal and ethical considerations, including copyright training and model usage policies, remain important as the technology evolves.Lire la suite

OpenAI has released ChatGPT Atlas, an AI‑infused web browser for macOS that embeds the ChatGPT assistant directly into the browsing experience. The browser offers features such as browser‑level memory, an experimental agent mode that can perform multi‑step tasks, and strong privacy controls, including incognito mode and parental supervision. While Windows, iOS and Android versions are slated for later release, Atlas aims to transform the web from a passive viewing platform into an active, task‑driven environment, positioning itself as a challenger to traditional browsers.Lire la suite