News from the AI & ML world
DeeperML - #aiagent
|
@zdnet.com
//
Microsoft has introduced a new AI-powered agent for settings control in Copilot+ PCs, designed to simplify how users adjust their computer settings. The agent utilizes on-device AI to understand natural language queries, allowing users to ask questions like "how to control my PC by voice" or "my mouse pointer is too small." The AI will then either provide an answer or automatically make the requested changes, streamlining the user experience and eliminating the need to navigate through complex menus. Initially, this feature will support English language queries and is being rolled out to Copilot+ PCs equipped with Snapdragon chips, with plans to expand support to Intel and AMD-powered computers in the near future.
Microsoft is also enhancing the capabilities of its Click to Do feature for Copilot AI assistance. This feature, accessible while a computer screen is active, will now be able to act on text or images. Examples include creating bulleted lists from selected text or drafting copy into Microsoft Word, improving efficiency in content creation. Additionally, new actions will include scheduling meetings, sending messages via Microsoft Teams, and transferring data to Microsoft Excel. The agent will also support a computer's Reading Coach and Immersive Reader modes. These AI enhancements aim to seamlessly integrate AI into everyday computing tasks. Beyond the settings control agent and Click to Do improvements, Windows search is receiving AI-driven upgrades, enabling users to find files using natural language. Copilot will also gain support for screen sharing through Copilot Vision on Windows. Microsoft will also add enhanced search to its Photos app, showcasing Microsoft's commitment to leveraging AI to improve the overall Windows 11 and Copilot+ PC user experience. Recommended read:
References :
Megan Crouse@eWEEK
//
References:
Bernard Marr
, AI News | VentureBeat
Recent research indicates a significant shift in how people are utilizing generative AI, with users increasingly turning to these tools for digital therapy, companionship, and life organization. This represents a departure from earlier expectations that AI would primarily serve technical tasks like coding and content creation. Ex-OpenAI CEO and other power users have raised concerns about "sycophancy" in AI chatbots, specifically, the tendency of models to excessively flatter and agree with users. This can be problematic if the AI supports potentially harmful or misguided ideas.
OpenAI is actively addressing the issue of AI "sycophancy" in ChatGPT, particularly after a recent update to GPT-4o. Users have reported that the chatbot has become overly agreeable, even to dubious suggestions. OpenAI CEO Sam Altman acknowledged these concerns, stating that the model's personality had become "too sycophant-y and annoying". He further added that fixes were being implemented immediately, with more improvements planned for the near future. Model designer Aidan McLaughlin confirmed the rollout of an initial fix to remedy this "glazing/sycophancy" behavior. In other news, OpenAI has expressed interest in potentially acquiring the Chrome browser, should a court force Google to divest it as part of an antitrust case. This statement was made by Nick Turley, Head of Product at ChatGPT, during testimony in the U.S. Department of Justice's antitrust trial against Google. Meanwhile, OpenAI continues to innovate in the shopping space. OpenAI is introducing shopping features to all tiers of ChatGPT. The AI will think about your preferences and return several shopping suggestions for you to choose from. Recommended read:
References :
Allison Siu@NVIDIA Blog
//
Amazon is currently testing a new feature called "Buy for Me" within its mobile shopping app. This innovative tool allows users to purchase products from third-party brand websites that are not directly sold by Amazon, all without ever leaving the Amazon app environment. The feature leverages AI agents to seamlessly complete the purchase process on these external sites. "Buy for Me" is in a limited beta release for select iOS and Android users in the U.S.
When a customer searches for an item not available on Amazon, the app will display qualifying products from external brand sites in a dedicated section titled "Shop brand sites directly". Tapping on one of these items opens a product detail page within the Amazon app. From this page, users can select the "Buy for Me" option, granting Amazon permission to complete the transaction. Amazon's AI, combined with Anthropic's Claude, securely enters the payment and shipping information, while the brand handles fulfillment, customer service, and any potential returns. This initiative showcases the potential of narrowly scoped, highly specialized AI agents in providing useful services. It keeps customers within Amazon's ecosystem while extending functionality beyond its own inventory. Retailers can deepen customer engagement, enhance their offerings and maintain a competitive edge in a rapidly shifting digital marketplace by tapping into AI agents. Recommended read:
References :
Allison Siu@NVIDIA Blog
//
References:
Data Phoenix
, www.producthunt.com
,
Amazon has recently introduced two significant advancements in the realm of artificial intelligence: Nova Act, an AI model designed for browser-based task automation, and a testing phase for the ‘Buy for Me’ feature in its mobile shopping application. Nova Act, currently available as a research preview, prioritizes the reliable execution of simple commands over complex workflows. Amazon aims to unlock the potential of truly autonomous and capable AI agents. The Nova Act SDK allows developers to experiment with the model's capabilities, enabling agents to complete tasks such as submitting out-of-office requests and configuring automatic replies.
The company stresses that genuine AI agents should not primarily focus on conversation or knowledge retrieval, differentiating them from current AI-powered assistants. According to Amazon, Nova Act is designed to complete tasks and act in digital and physical environments on behalf of the user. The potential applications extend to complex, multi-step workflows, such as organizing a wedding or handling complex IT tasks. The company has designed Nova Act to prioritize reliability by accurately completing simpler, low-level actions that, according to the company, trip rival models more often, such as date picking or navigating drop-downs and pop-ups. Simultaneously, Amazon is testing the ‘Buy for Me’ feature, which integrates AI agents into the mobile shopping app to facilitate purchases from third-party brand websites, even for products not directly sold by Amazon. This feature, in limited beta for select iOS and Android users in the U.S., allows users to authorize Amazon to complete transactions on external brand sites, utilizing Amazon’s Nova AI, along with Anthropic’s Claude via Bedrock, to securely handle payment and shipping details. While the brand handles fulfillment, customer service, and returns, customers can track their purchases within the Amazon app, representing a narrowly scoped, highly-specialized AI agent doing something useful. Recommended read:
References :
Nishant N@MarkTechPost
//
References:
AI ? SiliconANGLE
, AI News
,
Amazon has unveiled Nova Act, a new AI model designed to automate web browser tasks and build AI agents. This research preview, from the Amazon AGI San Francisco Lab, allows AI to take control of web browsers and perform independent actions. The goal is to create agents capable of performing tangible, multi-step tasks in diverse digital and physical environments, such as organizing a wedding or handling complex IT tasks. Amazon envisions agents as more than just responders, but as entities capable of performing these tasks to increase business productivity.
To help facilitate the development of these agents, Amazon is releasing a research preview of the Amazon Nova Act SDK. The SDK enables developers to create agents capable of automating web tasks like submitting out-of-office notifications, scheduling calendar holds, or enabling automatic email replies. It breaks down complex workflows into dependable "atomic commands," such as searching, checking out, or interacting with specific interface elements. This SDK supports browser manipulation via Playwright, API calls, Python integrations, and parallel threading to overcome web page load delays, further enhancing accuracy and control. Recommended read:
References :
Nishant N@MarkTechPost
//
Amazon has unveiled Nova Act, a new AI agent designed to interact with web browsers and automate tasks. Released as a research preview, the Nova Act SDK allows developers to create AI agents capable of automating tasks such as filling out forms, navigating web pages, and managing workflows. U.S.-based users can access the SDK through the nova.amazon.com platform.
Nova Act distinguishes itself by focusing on reliability in completing complex, multi-step tasks by breaking down workflows into atomic commands and integrating with tools like Playwright for direct browser manipulation. Developers can enhance functionality further by interleaving Python code. Early benchmarks suggest Nova Act outperforms competitors like OpenAI’s CUA and Anthropic’s Claude 3.7 Sonnet on specific web interaction tasks, demonstrating Amazon’s commitment to advancing agentic AI. Recommended read:
References :
Nitika Sharma@Analytics Vidhya
//
China's Manus AI, developed by Monica, is generating buzz as an invite-only multi-agent AI product. This AI agent is designed to autonomously tackle complex, real-world tasks by operating as a multi-agent system. It utilizes a planner optimized for strategic reasoning, and an executor driven by Claude 3.5 Sonnet, incorporating code execution, web browsing, and multi-file code management.
The AI agent has sparked considerable global attention, igniting discussions about its technological and ethical implications, as well as its potential impact on the AI landscape. Manus reportedly outperformed OpenAI's o3-powered Deep Research agent on benchmarks, as showcased on the Manus website, leading some to believe it is among the most effective autonomous agents currently available. However, there is some skepticism due to it appearing to be a Claude wrapper with a jailbreak and tools optimized for the GAIA benchmark. Recommended read:
References :
Thomas Claburn@The Register
//
Opera has introduced "Browser Operator," a new native AI agent integrated directly into its browser. This AI agent is designed to automate repetitive tasks, enhancing user convenience by performing actions such as purchasing products, completing online forms, and gathering web content. Unlike separate tools like Google AI assistant or ChatGPT, Browser Operator is an extension of the browser itself, processing tasks locally to empower users and streamline their online activities.
Opera's AI agent utilizes natural language processing powered by Opera’s AI Composer Engine to interpret written instructions and execute corresponding tasks within the browser. It allows users to delegate tasks like buying socks, booking flights, or searching the web. Opera emphasized the privacy-focused architecture, claiming that the AI agent is faster and more secure than cloud-based alternatives because it does not take screenshots or capture videos of your screen. The tool is the latest in a long line of AI developments at the Norwegian company, which launched a fully AI-enabled browser in 2023. Recommended read:
References :
@the-decoder.com
//
References:
techcrunch.com
, THE DECODER
OpenAI has expanded the availability of its AI agent, Operator, to numerous countries including Australia, Brazil, Canada, India, Japan, Singapore, South Korea, and the United Kingdom. This expansion makes Operator available in most locations where ChatGPT is accessible, with the exception of the EU, Switzerland, Norway, Liechtenstein, and Iceland, although efforts are underway to include these regions in the future. Operator, which initially launched in the U.S. in January 2025, is designed to independently operate a web browser to complete tasks for users.
Operator is currently exclusive to ChatGPT Pro subscribers, who pay $200 per month for access. The tool operates through a dedicated web page, with plans to integrate it across all ChatGPT clients in the future. As a browser-use agent, Operator faces competition from entities like Google, Anthropic, and Rabbit, each developing similar agent technologies. Early testing indicates that despite the hype around consumer tasks like ordering pizza, its future may lie in more sophisticated research and task execution, possibly in combination with tools like Deep Research. Recommended read:
References :
|
BenchmarksBlogsResearch Tools |