AI News

News

Page 63
AI Models Store Memories and Reasoning in Distinct Neural Regions

AI Models Store Memories and Reasoning in Distinct Neural Regions

Researchers have found that artificial intelligence models keep memorized facts and reasoning abilities in separate parts of their neural networks. By analyzing the loss landscape, they discovered that memorized items create sharp spikes while reasoning produces smoother curves. The study also explored early techniques for removing specific data from models, noting that complete elimination cannot yet be guaranteed. These insights could guide future efforts to manage and protect sensitive information in AI systems.

Google’s Gemini Voice Assistant Rolls Out Slowly Amid Early Glitches

Google’s Gemini Voice Assistant Rolls Out Slowly Amid Early Glitches

Google has begun rolling out its new Gemini voice assistant for Google Home smart speakers and displays, positioning it as the default assistant across its ecosystem. While a recent blog post praised the rollout and highlighted a user’s positive experience, the accompanying FAQ revealed recurring problems, including misinterpreted commands and unreliable device control. Access to Gemini requires enrollment in an Early Access program that launched in late October, and the rollout is expected to be gradual, potentially not reaching all users until next spring. Early feedback suggests the generative‑AI‑driven assistant still struggles with basic smart‑home tasks.

Google Brings Gemini AI to Google TV Streamer, Replacing Assistant

Google Brings Gemini AI to Google TV Streamer, Replacing Assistant

Google announced that its Gemini generative AI is being rolled out to the Google TV Streamer, where it will replace Google Assistant. The new system lets users ask natural language questions about movies, TV shows, education, DIY projects and more, using the remote microphone button. The rollout begins this week for users aged 18 and older and will expand to additional streaming devices such as select TCL and Hisense models and the Walmart Onn 4K Pro. The move is part of Google’s broader plan to shift its voice ecosystem from Assistant to Gemini across all platforms.

Tech Giants Bet Billions on AI Infrastructure Amid Uncertain Demand

Tech Giants Bet Billions on AI Infrastructure Amid Uncertain Demand

Major technology firms are committing unprecedented sums to AI‑related data center projects, with Oracle, SoftBank, Meta and others pledging hundreds of billions of dollars. While the scale of funding reflects optimism about AI growth, surveys show many companies remain cautious, and industry leaders warn that physical infrastructure—especially power and space—could become a bottleneck. The juxtaposition of massive financial commitments and lingering market uncertainty highlights the complex dynamics shaping the emerging AI ecosystem.

Gamma Secures $68 Million Series B, Hits $2.1 B Valuation and $100 Million ARR

Gamma Secures $68 Million Series B, Hits $2.1 B Valuation and $100 Million ARR

Gamma, the AI‑powered platform that creates presentations, websites and social media posts, announced a $68 million Series B round that lifts its valuation to $2.1 billion. Led by Andreessen Horowitz, the round also includes participation from Accel, Uncork Capital and other existing investors. Co‑founder and CEO Grant Lee said the company has reached $100 million in annual recurring revenue on a profitable basis and now serves roughly 70 million users. The funding will support further product development and global expansion.

Wikipedia Calls on AI Companies to Use Paid API and Halt Scraping

Wikipedia Calls on AI Companies to Use Paid API and Halt Scraping

The Wikimedia Foundation announced that it expects AI developers to access Wikipedia content through its paid Wikimedia Enterprise platform rather than scraping the site. The move aims to ensure proper attribution, reduce server strain, and support the nonprofit’s mission as human page views decline. The foundation emphasized that generative‑AI tools should credit Wikipedia contributors and direct users back to the source, warning that continued scraping could impact volunteer engagement and donor support.

Study Highlights Optimal US Sites for AI Data Centers Amid Environmental Concerns

Study Highlights Optimal US Sites for AI Data Centers Amid Environmental Concerns

A new analysis published in Nature Communications examines the environmental footprint of the rapid AI‑driven data‑center expansion in the United States. While tech giants pour billions into new facilities, the study finds that location dramatically influences carbon and water impacts. States such as Texas, Montana, Nebraska and South Dakota emerge as optimal sites because they balance lower water scarcity with cleaner electricity grids. In contrast, traditional hubs like Virginia and California face growing pressure on energy and water resources, raising doubts about the feasibility of industry net‑zero pledges.

Google Brings Gemini AI Directly Into Photos, Turning Personal Libraries into Instant Answer Machines

Google Brings Gemini AI Directly Into Photos, Turning Personal Libraries into Instant Answer Machines

Google has integrated its Gemini generative‑AI model into Google Photos via a new “Ask” button. The feature lets users query their own photo collections—rather than the web—for specific images, delivering results in seconds. A user with a library of nearly 172,000 photos demonstrated the tool by asking for pictures of a bay window with shades, receiving dozens of relevant shots and even contextual commentary. The rollout showcases Google’s push to embed Gemini across its apps and highlights how AI can streamline everyday tasks like locating a single photo in a massive personal archive.

Google Maps Unveils New Gemini-Powered AI Tools for Developers and Users

Google Maps Unveils New Gemini-Powered AI Tools for Developers and Users

Google Maps announced a suite of AI-driven features powered by Gemini models, aimed at both developers and everyday users. The rollout includes a Builder Agent that generates map‑based prototypes from text prompts, a styling agent for custom map designs, and a Model Context Protocol (MCP) server that links AI assistants to Maps documentation. Additional tools such as Grounding Lite and Contextual View provide low‑code visual answers to location queries. The enhancements also extend to consumer experiences, with hands‑free Gemini navigation and region‑specific alerts in select markets.

OpenAI’s Sora App Tops Google Play Store’s Free Chart

OpenAI’s Sora App Tops Google Play Store’s Free Chart

OpenAI’s AI‑video platform Sora has surged to the No. 1 spot in the Google Play Store’s free‑app rankings, overtaking its sister app ChatGPT. The Android version logged 470,000 downloads on its first day, far exceeding the initial iOS launch numbers. New creation tools, cameo features, basic editing, and expanded social options are rolling out, while payment plans now let users purchase extra video generations for $4. Pro users gain longer video limits and storyboarding capabilities. OpenAI’s head of Sora, Bill Peebles, outlined these updates and hinted at further professional‑grade enhancements.

1Mind Raises $30 Million to Expand AI Sales Agent Mindy

1Mind Raises $30 Million to Expand AI Sales Agent Mindy

AI sales startup 1Mind announced a $30 million Series A round led by Battery Ventures, bringing its total funding to $40 million. Co‑founder Amanda Kahlow, formerly of 6Sense, said the company’s AI agent, named Mindy, focuses on inbound sales and can handle tasks from web‑site assistance to closing enterprise deals. Mindy, built on a mix of large‑language models including OpenAI and Google Gemini, uses deterministic AI to limit hallucinations and is designed to say when it does not know an answer. More than 30 companies, including HubSpot, LinkedIn and New Relic, are already using Mindy, and the startup plans to grow its team and product capabilities.

AI Companies Face Growing Copyright Lawsuits Over Training Data

AI Companies Face Growing Copyright Lawsuits Over Training Data

AI developers are under increasing legal pressure as creators allege that large language model and image generator firms have used copyrighted works without permission to train their systems. More than thirty lawsuits have been filed against firms such as OpenAI, Google, Anthropic, and Meta, while the industry pushes for a fair‑use exemption to keep development costs low. Courts have delivered mixed rulings, with some judges deeming the use "exceedingly transformative" and others allowing settlements. The dispute highlights a clash between the need for rapid AI innovation and the protection of creators’ rights.