OpenAI Adds Real-Time Search

Sign Up | Advertise

The AI World Today

• OpenAI Adds Real-Time Search

• New Tool Prevents AI Failures

• KKR, ECP Invest in AI Infrastructure

  +

• Heads Up

• AI Solution

OpenAI Challenges Google with Real-Time ChatGPT Search

OpenAI has integrated a real-time search tool into ChatGPT, enhancing responses to prompts on sports, news, and weather, which places it in closer competition with Google and partner Microsoft’s Bing. Available from today for ChatGPT Plus and Team subscribers, the feature will reach enterprise and educational users in coming weeks and free users in the coming months. Users can access search by clicking a globe icon, and ChatGPT itself may route relevant queries for real-time information. Results, similar to regular ChatGPT responses, can now include maps, stock charts, and images. OpenAI currently has no plans for ads in search results, and its news sources include partners like Axel Springer and The Associated Press, among others.

Patronus AI Introduces Real-Time AI Monitoring Platform

Patronus AI, a San Francisco startup, has launched a self-serve platform to detect and prevent AI failures in real-time, backed by $17 million in Series A funding. Designed to address AI issues like hallucinations and security flaws, the platform introduces customizable “judge evaluators,” allowing companies to create rules in plain English. Core to its system is Lynx, a hallucination detection model outperforming GPT-4 by 8.3% in detecting medical inaccuracies. Additional tools, including CopyrightCatcher and FinanceBench, provide specialized error-checking. With a pay-as-you-go pricing model starting at $10 per 1,000 API calls, Patronus AI aims to make AI safety accessible to businesses of all sizes, gaining early clients such as HP, Nvidia, and Pearson.

KKR and ECP Commit $50B to AI Infrastructure

Private equity firms KKR and Energy Capital Partners (ECP) announced a $50 billion partnership to invest in data centers and energy infrastructure to support AI’s growing demands. This partnership focuses on the essential architecture that will power AI technologies across industries. According to Goldman Sachs, data center electricity demand could rise by 160% by 2030, with McKinsey estimating up to a 300% increase. KKR will fund the initiative through existing infrastructure and real estate funds, as well as managed insurance accounts, while ECP will leverage both existing and future infrastructure funds. This move represents a shift in private equity, as firms now invest indirectly to bolster startup and AI industry growth from the ground up.

Heads Up 

OpenAI held its first international Dev Day in London, unveiling new voice API capabilities and hosting a Reddit AMA with CEO Sam Altman and leadership.

Nvidia's new research introduces HOVER, a compact 1.5M parameter neural network enabling versatile, whole-body robotic control across different movement modes and input methods.

Google Maps, powered by Gemini AI, is adding features to help U.S. users explore new places and get answers about various locations starting Thursday.

Meta is partnering with GelSight and Wonik Robotics to develop tactile sensors for AI, aimed at helping scientists advance AI's understanding of the physical world. 

Chinese research institutions tied to the People’s Liberation Army have reportedly adapted Meta’s open-source Llama model to create AI tools for military applications.

AI video platform D-ID has launched new Express and Premium+ avatars, aiming to enhance humanlike video creation for enterprise use in marketing, sales, and customer support.

Anthropic's Claude is now accessible on Apple and Windows desktop apps, with new dictation features launching for mobile and iPad users.

AI Solution

Google Enhances Gemini API with Real-Time Search Grounding

Starting today, developers using Google’s Gemini API and AI Studio can ground their prompts with real-time data from Google Search, enhancing accuracy with up-to-date information. Available on a paid tier ($35 per 1,000 queries), grounding enables models to connect with Google’s search catalog, improving responses and reducing hallucinations. Developers can toggle grounding on and set its frequency through “dynamic retrieval” options to match their needs for current or detailed information. AI Studio’s compare mode lets users contrast grounded results with standard model responses. This feature also includes citation links, offering transparency and source credit for enriched answers. Google’s AI Studio now supports developers from testing Gemini models to directly generating code, streamlining AI project development.