Category: AI

Google has introduced a new feature that allows developers to integrate live Google Maps data into applications powered by its Gemini AI models. This integration, detailed on Google’s blog, enables developers to leverage over 250 million places’ data for more intelligent and contextually rich experiences.

With this integration, applications can provide detailed information based on the user’s location, such as business details, reviews, and venue atmosphere. This capability is particularly beneficial for sectors like local search, delivery services, real estate, and travel planning, where location plays a crucial role.

Developers using Google’s AI Studio can access this feature through the Gemini Live API, supporting models like Gemini 2.5 Pro, Flash, and Flash-Lite. The seamless integration of Maps data enhances response quality and accuracy, enabling developers to create more informed and location-aware applications.

By combining Google Maps data with the Gemini API, developers can generate contextually relevant responses, offering users a more personalized and detailed experience. The tool supports various use cases across industries, from itinerary generation to personalized local recommendations.

Customization options allow for tailored interactions, while the API’s structured metadata ensures transparency and trust in user-facing applications. Google’s emphasis on proper attribution of Maps-based sources further enhances credibility and user trust.

Grounding with Google Maps, now available globally, marks an advancement in the Gemini API’s capabilities, empowering developers to build AI-driven applications that better understand and respond to real-world contexts.

Source: VentureBeat

Adobe has introduced a new model customization service called Adobe AI Foundry to cater to enterprise teams, as reported by VentureBeat. This service aims to create tailored versions of its flagship AI model, Firefly, for businesses. The Adobe AI Foundry service involves rearchitecting and retraining Firefly models specific to each client, offering a more sophisticated approach compared to traditional custom models.

Unlike custom models that focus on a single concept, Foundry models comprehend multiple concepts and are multimodal, enabling a broader range of applications beyond image processing. By understanding a company’s brand tone, image style, products, and services, the models generate content aligned with the client’s requirements.

Adobe’s Vice President of GenAI New Business Ventures, Hannah Elsakr, explained that the AI Foundry service involves deep tuning, going beyond simple fine-tuning to extensively modify base models. This process includes incorporating enterprise intellectual property into the models while keeping it separate and owned by the client.

The AI Foundry service will be delivered through Adobe’s API solution, Firefly Services, and involves collaboration with enterprise customers to enhance model training and customization. Early adopters of this service include Home Depot and Walt Disney Imagineering, showcasing the appeal of advanced AI model customization in diverse industries.

Source: VentureBeat

Anthropic, a leading provider of AI-powered coding services, has announced the launch of Claude Code, a platform that enables developers to execute parallel coding tasks. This innovation marks a significant advancement in Anthropic’s Vibe coding capabilities, offering enhanced asynchronous functionality. Previously accessible through PCs, Claude Code is now available on the web and in preview on the Claude iOS app.

Developers can seamlessly connect their GitHub repositories and describe their coding needs, with Claude handling the implementation in isolated environments and providing real-time progress tracking. This shift towards asynchronous coding aligns with the demands of many enterprises seeking efficient workflows.

One of the key features of Claude Code is its ability to run multiple tasks in parallel across different repositories. This facilitates faster project completion, with automatic pull request creation and clear change summaries. Comparable platforms like Google’s Jules and Code Assist, as well as OpenAI’s Codex, also offer parallel tasking capabilities.

To ensure security, Claude Code operates in isolated sandbox environments with network and filesystem restrictions. Interactions are secured through a proxy service, limiting access to authorized repositories. Enterprise users have the flexibility to customize connectivity permissions.

Anthropic’s move to bring Claude Code to the web enhances accessibility, with the mobile version set to attract users interested in coding on-the-go. Powered by Claude Sonnet 4.5, known for its robust coding capabilities, Claude Code continues to innovate in the AI-powered coding landscape.

Source: VentureBeat

Recent research from MIT Sloan School of Management, as highlighted by VentureBeat, reveals that AI technology, particularly in the form of AI-powered PCs, is enhancing human creativity in the workplace and revolutionizing how tasks are approached and executed.

AI-powered PCs, equipped with Neural Processing Units (NPUs) and local AI processing, are empowering knowledge workers by providing the speed, security, and creative potential needed to drive innovation. These next-generation laptops are not merely tools for efficiency but catalysts for fresh ideas, quicker iteration, and more engaged teams across various departments.

The impact of AI-powered PCs is already evident in real-world scenarios. Marketing teams are producing campaign assets at an unprecedented pace, engineers are accelerating design cycles, and sales representatives are personalizing proposals on the fly. These advancements are not just about speed; they are about infusing creativity into everyday workflows, ultimately translating into tangible business outcomes such as faster time-to-market and enhanced customer engagement.

Despite the early successes, widespread adoption of AI-powered PCs for creative tasks remains limited among knowledge workers. HP’s research underscores the importance of providing employees with tools that foster creativity, leading to increased productivity, satisfaction, and retention rates. The divide in adoption presents an opportunity for CIOs to deploy advanced technology and cultivate a culture where creativity drives substantial business value.

By leveraging on-device AI capabilities, employees can work seamlessly without interruptions, regardless of connectivity constraints. The potential business outcomes span across marketing, product development, and sales, showcasing the multifaceted benefits of integrating AI-powered PCs into organizational workflows.

AI-powered PCs are not just about enhancing performance; they are about redefining how work is approached and experienced. By embracing AI technology that sparks creativity alongside productivity, organizations can propel innovation, engagement, and retention to new heights, setting a new standard for workplace dynamics.

Source: VentureBeat

OpenAI has announced the launch of ChatGPT Atlas, an AI-powered web browser aiming to challenge the dominance of Google Chrome. Available globally on Apple’s macOS and soon on Windows, iOS, and Android, Atlas seeks to revolutionize how users interact with the web.

The browser offers a seamless experience, allowing users to ask questions and chat with agents directly within the interface. By enabling search via prompts and questions, Atlas aims to enhance productivity and user experience. A key feature is the ability to call on agents for various tasks, initially catering to ChatGPT Business, Plus, and Pro users.

Unlike traditional browsers, Atlas integrates a chat function that personalizes user interactions over time. Users can engage with ChatGPT across different websites, leveraging its memory base to provide relevant responses. Additionally, Atlas streamlines tasks like email composition by allowing users to interact with ChatGPT within the browser, reducing the need for copy-pasting.

With the integration of agents for task automation, Atlas empowers users to accomplish various activities seamlessly. OpenAI’s emphasis on agent infrastructure signals a shift towards browser-integrated agents for enhanced user convenience. As AI models and chat platforms gain popularity for web searches, the introduction of AI-enabled browsers intensifies competition in the tech landscape.

Source: VentureBeat

Researchers at Mila have introduced a novel technique called Markovian Thinking, which aims to revolutionize the efficiency of large language models (LLMs) in complex reasoning tasks. This innovative approach, detailed in a recent paper, enables LLMs to engage in extended reasoning without the exorbitant computational costs that typically accompany such processes.

The key concept behind Markovian Thinking is the restructuring of the reasoning chain into fixed-size chunks within an environment known as Delethink. By breaking down the scaling issue that affects lengthy LLM responses, this method has demonstrated the potential to reduce training costs significantly, with initial estimates suggesting up to a two-thirds decrease for a 1.5B parameter model compared to traditional approaches.

One of the primary challenges that this technique addresses is the quadratic growth problem associated with long-chain reasoning in LLMs. By separating the model’s thinking duration from the amount of context it processes, the Markovian Thinker paradigm transforms the quadratic computational cost into linear requirements, offering a more efficient and effective approach to LLM reasoning.

Through Delethink, the model reasons in sequential fixed-size chunks, each containing 8,000 tokens, maintaining a constant reasoning context window. As a result, the AI learns to embed a summary of its progress, ensuring continuity in its reasoning across different chunks without modifying the original input prompt.

The implications of this novel approach are significant, particularly in enterprise applications where efficiency and performance are paramount. By enabling models to reason for longer durations with reduced computational costs, Markovian Thinking paves the way for next-generation AI capabilities, potentially unlocking the ability for models to ‘think’ for millions of tokens and driving advancements in scientific discovery.

Source: VentureBeat

DeepSeek, a Chinese AI research company, has unveiled the DeepSeek-OCR model, which sets a new standard for text compression. By treating text as images, DeepSeek’s model achieves a significant breakthrough, compressing text up to 10 times more efficiently than traditional methods. This innovation challenges the fundamental assumptions in AI development, potentially enabling language models with significantly expanded context windows.

The model’s architecture features the innovative DeepEncoder and a language decoder with 570 million activated parameters. DeepSeek’s approach offers unprecedented compression ratios while maintaining high OCR precision, enabling the processing of 200,000 pages per day on a single GPU and 33 million pages daily with a cluster of servers.

DeepSeek’s open-source release of the model weights and code has sparked industry interest, prompting speculation about potential proprietary techniques utilized by other AI labs. The model’s implications extend beyond text compression, suggesting a fundamental rethinking of how language models process information, with potential applications for reasoning over large visual tokens.

Source: VentureBeat

PayPal, a key player in the e-commerce realm, is introducing innovative features to enhance discoverability and flexibility for enterprises in the agentic commerce space. The company’s new discoverability solution allows businesses to seamlessly offer their products on various chat platforms, transcending the limitations of different payment protocols. By leveraging its participation in Google’s Agent Payments Protocol (AP2), PayPal is facilitating a smoother transition for merchants into agentic commerce, emphasizing the importance of adaptability in the evolving landscape of AI-powered shopping.

Michelle Gill, PayPal’s General Manager for small business and financial services, highlighted the significance of early preparation for AI-driven shopping experiences. The introduction of Agent Ready and Shop Sync under PayPal’s Agentic Commerce Services further underscores the company’s commitment to providing practical solutions for merchants. Agent Ready enables existing PayPal merchants to accept payments on AI platforms, while Shop Sync ensures the seamless integration of product data across various AI chat interfaces.

As the agentic commerce ecosystem continues to evolve, the benefits of PayPal’s services become evident. These include fast integration with partners, enhanced product discovery experiences, and the preservation of customer relationships through controlled communications and insights. Despite being currently accessible only through Perplexity, PayPal intends to expand its reach to more platforms in the near future, catering to the diverse needs of businesses operating in the AI commerce space.

Amidst the fragmented landscape of AI platforms vying for standardization in transactions, PayPal’s focus on flexibility stands out as a strategic advantage. Gill emphasized the challenge faced by enterprises in choosing where to invest their integration efforts, given the multitude of protocols and models offered by different platforms. The company’s approach underscores the importance of adaptability and foresight in navigating the complexities of agentic commerce.

Source: VentureBeat

Intuit, known for its financial software, has unveiled Intuit Intelligence, a system integrating AI agents into its QuickBooks platform for tasks like sales tax compliance and payroll processing. These agents, part of the GenOS update, enhance accounting capabilities and offer a unified interface for querying data across various sources with natural language. Despite advancements, Intuit faced challenges as even significant accuracy improvements led to customer trust issues in finance.

Unlike relying on large language models, Intuit prioritizes real data queries for financial insights. By accessing diverse data sources like third-party systems and user uploads, QuickBooks ensures reliable AI operations. This design choice minimizes the risk of AI hallucinations and shadow AI usage, common in financial settings.

Explainability is central to Intuit’s AI design, displaying the reasoning behind automated decisions to users. This transparency fosters trust, crucial in a context where errors erode customer confidence. By integrating AI agents into existing workflows, Intuit eases users into conversational interfaces without abrupt changes.

Lessons from Intuit’s AI journey offer valuable insights for enterprise AI initiatives. Prioritizing trustworthiness over flashy capabilities, focusing on accuracy, transparency, and human oversight sets a strong foundation for successful AI deployment in critical domains.

Source: VentureBeat

Researchers from Meta FAIR and the University of Edinburgh have unveiled a new technique called Circuit-based Reasoning Verification (CRV) that enhances the accuracy of large language models (LLMs) by identifying and correcting reasoning errors. By monitoring internal ‘reasoning circuits’ within LLMs, CRV can pinpoint computational mistakes and intervene to correct faulty reasoning in real-time. This innovation addresses a significant challenge in AI by ensuring the fidelity and correctness of model reasoning, crucial for deploying reliable AI applications in the enterprise sector.

The CRV approach revolves around investigating chain-of-thought (CoT) reasoning, a method used to enhance LLM performance on complex tasks. While CoT has been effective, its reliability has been questioned due to flawed reasoning processes. Existing verification methods primarily rely on ‘black-box’ and ‘gray-box’ approaches but lack the ability to explain why computation failures occur, posing limitations for real-world applications.

CRV adopts a white-box approach by making LLMs interpretable through the use of ‘transcoders,’ which transform internal computations into meaningful features for diagnostic analysis. By constructing attribution graphs and ‘structural fingerprints’ for each reasoning step, CRV trains a ‘diagnostic classifier’ to predict correctness. Empirical tests on a modified Llama 3.1 8B Instruct model demonstrated CRV’s superiority over conventional methods, showcasing its domain-specific error detection capabilities and causal understanding of reasoning failures.

The potential impact of CRV extends beyond research, offering insights into developing AI model debuggers that can precisely identify and rectify reasoning errors. This advancement could lead to more robust LLMs and autonomous agents capable of self-correction, enhancing their adaptability to real-world challenges and reducing the need for extensive retraining.

Source: VentureBeat

Canva, known for its design platform, is integrating AI to facilitate a new era of creativity and collaboration. Canva’s co-founder and CPO, Cameron Adams, emphasizes the potential of AI to empower individuals and enterprises in turning imagination into reality. The company’s Creative Operating System (COS) incorporates AI at every level of content creation, offering a comprehensive platform beyond traditional design tools.

Canva’s COS is structured as a three-layer stack, with AI playing a crucial role in enabling real-time creation and collaboration. The platform’s AI-driven functionalities allow users to seamlessly design, edit, and launch diverse content types within a unified dashboard, enhancing efficiency and creativity.

New features like ‘Ask Canva’ provide users with design advice and editing capabilities, further streamlining the creative process. Canva’s strategic focus on automation and intelligence aligns with the growing demand for personalized and consistent branding across various platforms.

With over 250 million users and notable clients like Walmart and Disney, Canva’s impact on design and marketing workflows is evident. Its COS has transformed content creation, saving time and resources for enterprises while enhancing collaboration between teams.

Competing in a landscape with tools like Adobe Express and Figma, Canva stands out for its extensive template library, user-friendly interface, and broad range of content creation options. The platform’s pricing tiers cater to individual users, teams, and enterprises, offering flexibility for diverse needs.

Canva’s commitment to human-AI collaboration and openness sets a precedent for technology teams, emphasizing the symbiotic relationship between people and technology. By working with top AI providers and developing proprietary models, Canva exemplifies a forward-looking approach to innovation and creativity.

Source: VentureBeat

OpenAI has introduced Aardvark, a GPT-5-powered autonomous security agent designed for code analysis and patching. Available in private beta, Aardvark mimics human expert vulnerability identification processes with its multi-stage, LLM-driven approach for continuous code analysis, exploit validation, and automated patch generation. Serving as a scalable defense tool, Aardvark has demonstrated high recall and effectiveness in identifying both known and synthetic vulnerabilities, enhancing security in modern software development environments.

Operating as an agentic system, Aardvark utilizes LLM reasoning to interpret code behavior and uncover vulnerabilities. Its structured process includes threat modeling, commit-level scanning, validation sandbox, and automated patching, seamlessly integrating with GitHub and Codex to provide non-intrusive security scanning. Aardvark’s performance metrics showcase its ability to identify a significant percentage of issues with a low false positive rate, demonstrating its potential as a proactive security solution.

The release of Aardvark aligns with OpenAI’s strategic focus on specialized AI agents targeting specific capabilities within real-world environments. By offering a comprehensive security solution, Aardvark could revolutionize how security is embedded in continuous development environments, providing a force multiplier for security teams and AI engineers. Its integration capabilities with modern AI operations stacks and data infrastructure tools position Aardvark as a valuable addition to enhancing security checks while maintaining development agility.

Source: VentureBeat