Category: Artificial intelligence

  • China Develops Military AI Tool Using Meta’s Llama Technology

    China Develops Military AI Tool Using Meta’s Llama Technology

    Chinese researchers associated with the People’s Liberation Army have developed an AI tool designed to collect and assess data, aiming to deliver accurate information for military operations.

    Tool Overview

    A report from Reuters reveals that the tool, named "ChatBIT," is built on an earlier version of the publicly accessible Llama 13B LLM (Large Language Model) created by Meta. In a research paper published in June, the scientists mentioned that ChatBIT had been "optimized for dialogue and question-answering tasks in the military domain."

    Policy Conflict

    This development directly contradicts Meta’s Acceptable Use Policy, which forbids the application of its model for "Military, warfare, nuclear industries or applications, espionage, use for materials or activities that are subject to the International Traffic Arms Regulations (ITAR) maintained by the United States Department of State."

    In a communication to Reuters, Molly Montgomery, who is Meta’s director of public policy, stated that any utilization of their model by the People’s Liberation Army is unauthorized and goes against their acceptable use policy.

    Response from Meta

    Another representative from Meta commented that an outdated "version of an American open-source model is irrelevant when we know China is already investing more than a trillion dollars to surpass the US on AI." This highlights the ongoing competition between the two nations in the field of artificial intelligence.

  • Microsoft Postpones Copilot+ Recall to December

    Microsoft Postpones Copilot+ Recall to December

    Microsoft has once again postponed the Recall feature for Copilot+ PCs. This feature is designed to simplify the process of finding items on Windows by capturing screenshots of user activity at set intervals, which allows users to search through this information. It employs local AI models to create a timeline for users to explore.

    Initial Plans and Delays

    Originally, Microsoft intended to launch this feature in June. However, due to pushback from users and security experts, the rollout was pushed back to November. The company also decided that Recall would first be available to Microsoft Insiders for testing rather than being offered as an opt-in feature as previously planned.

    Further Postponement

    Now, Microsoft has announced yet another delay, moving the launch to December. Brendon LeBlanc, who is a senior product manager for Windows, informed The Verge that Microsoft is "dedicated to providing a secure and trustworthy experience with Recall" and is "taking more time to enhance the experience before introducing it to Windows Insiders".

    Hardware Requirements

    Recall will only be available on "Copilot+ PCs," which are computers equipped with dedicated NPUs (Neural Processing Units). To be classified as a "Copilot+ PC," Microsoft requires an NPU that can perform 40 trillion operations per second, along with 16GB of RAM and 256GB of solid-state storage.

    The Verge, Microsoft Windows Blogs.

  • Meta’s Llama 4 Uses 100,000 Nvidia H100 GPUs for Training

    Meta’s Llama 4 Uses 100,000 Nvidia H100 GPUs for Training

    Meta has just announced a new update on Llama 4, their upcoming language model. During a recent earnings call, CEO Mark Zuckerberg revealed that they are training Llama 4 using a setup with over 100,000 Nvidia H100 GPUs. This is a larger setup than anything previously reported by competitors.

    Upcoming Features of Llama 4

    This new language model is set to launch in early 2025, starting with its smaller versions first. While specific capabilities are not fully disclosed, Zuckerberg suggested that Llama 4 will have enhanced features, improved reasoning abilities, and will operate at a quicker pace overall.

    Meta’s Unique Strategy

    Meta continues its strategy of offering models for free download, unlike OpenAI and Google, which restrict access through APIs. This makes Llama 4 particularly appealing for startups and researchers who prefer more flexibility in using AI technologies.

    Significant Energy and Financial Implications

    Given the extensive computing resources, the energy requirements are also considerable, estimated at around 150 megawatts—this is five times more than what the largest supercomputer at a U.S. national lab consumes. Meta’s financial plan reflects this ambitious scale, with infrastructure expenditures projected to hit $40 billion in 2024, marking a 42 percent increase from 2023.

    Despite these hefty expenses, Meta’s financial health remains strong, showing a 22 percent rise in sales, primarily driven by advertising revenue, which has helped offset a 9 percent increase in operating costs.

  • Google Play Store to Introduce AI-Powered Ask a Question Feature

    Google Play Store to Introduce AI-Powered Ask a Question Feature

    Google is said to be developing an AI-powered tool for the Play Store that aims to assist users with questions about apps and help them find new ones. An analysis of the APK by Android Authority revealed that Google Play Store version 43.3.32-31 includes new strings labeled "Ask a Question," which appears to be the name of this upcoming feature.

    AI-Generated FAQs

    Previously, the publication shared details about AI-created FAQs for the Play Store. This feature automatically generates a FAQ (Frequently Asked Questions) section on app listings, providing users with essential information about an app.

    User Interaction

    The report indicates that this feature will likely show up as a button on both app listings and search results. When users click the button, they can type in their questions, and the AI will respond with answers. In theory, this could enhance the process of discovering new apps. However, Google has not made any official announcement regarding this feature, and it is not currently available on the Play Store. It’s also important to mention that not every feature found in the code makes it into the final version.

    Android Authority

    Image Source: Photo by Mika Baumeister on Unsplash

  • OpenAI and Broadcom Team Up to Create AI Inference Chip

    OpenAI and Broadcom Team Up to Create AI Inference Chip

    OpenAI has been reported to have teamed up with Broadcom, a US-based chip design company, to develop a specialized AI inference chip. Insiders from OpenAI shared with Reuters that the organization is aiming to broaden its supply chains to help lower costs.

    Shift in Manufacturing Plans

    Previously, OpenAI had ambitions to create a network of foundries to produce chips internally. However, those ambitions are currently paused due to the high expenses and the lengthy duration required to establish them.

    New Partnerships and Strategies

    As OpenAI works towards an in-house chip, it has started integrating AI chips from AMD and Nvidia into its operations. The sources mentioned that the company plans to utilize collaborations with other companies to manage chip supplies, combining both internal and external resources.

    Team Expansion and Future Production

    Two insiders informed Reuters that OpenAI is evaluating its chip design components and might bring on additional external partners. The organization has assembled a team of about 20 individuals, which includes Thomas Norrie and Richard Ho, who have prior experience working on TPUs (Tensor Processing Units) at Google.

    OpenAI has successfully arranged for manufacturing capabilities with TSMC (Taiwan Semiconductor Manufacturing Company) thanks to Broadcom’s assistance and aims to start the production of its in-house chips by 2026. However, this schedule may be subject to adjustments, according to the insiders.

    Reuters

    Image Source: OpenAI

  • Linus Torvalds Calls AI Industry 90% Marketing

    Linus Torvalds Calls AI Industry 90% Marketing

    The creator of the Linux kernel, Linus Torvalds, isn’t exactly excited about the current state of artificial intelligence, expressing that it seems more focused on hype rather than practical applications.

    Thoughts on AI’s Current Landscape

    During his talk at the Open Source Summit in Vienna, Torvalds remarked that the present AI environment feels like “90 percent marketing and 10 percent reality.” He acknowledged that while machine learning and generative AI hold potential for the future, he wasn’t afraid to criticize the surrounding excitement.

    Waiting for Maturity in AI

    Torvalds expressed that he prefers to wait until the technology is more developed before he embraces it. He anticipates that AI will become clearer in its usefulness over the next few years, with machine learning and generative algorithms gaining more significance and practicality. He believes that soon we will understand better what AI can do to simplify our daily tasks.

    Acknowledgment of Useful Tools

    While he recognized that tools such as ChatGPT are impressive demonstrations of technology and have their applications, especially in fields like graphic design, he consistently returned to the notion that the entire AI sector feels excessively glorified by flashy promotions and inflated market values, with Nvidia being a notable example.

    His viewpoint is consistent with recent comments made by other figures in technology, such as the CEO of Baidu, who forecasted that 99 percent of the existing AI companies would eventually fail.

  • Meta Developing New AI-Powered Search Engine

    Meta Developing New AI-Powered Search Engine

    Meta is working on creating its own AI-driven search engine that will offer answers and summaries about current events and individuals through its Meta AI chatbot. At the moment, this feature depends on Google Search and Microsoft’s Bing for information.

    Reducing Dependence on Major Players

    According to a report from The Information (via Engadget), Meta’s initiative aims to lessen its reliance on Google and Microsoft. A source who wished to remain anonymous mentioned that Meta is looking for a backup solution in case either Google or Microsoft decides to end their partnership in the future.

    New Partnership for Real-Time Updates

    Recently, Meta entered into a multi-year agreement with Reuters, which lets it present real-time news updates to users in the United States through the chatbot. A few months back, Business Insider (via The Verge) shared that Meta’s web crawler was collecting data to train its AI systems.

    Independent Data Indexing

    The Information confirmed this information through its contacts within Meta, who stated that the company has been gathering and indexing data for the past eight months. The strategy behind this move is to make Meta as self-sufficient as possible and minimize its reliance on other major tech companies.

    The Information, Engadget

    Image Credit: By Petar Adžaga – Own work, CC BY-SA 4.0, Link

  • Stable Diffusion Models for Local Hosting on Google Pixel 10 Pro

    Stable Diffusion Models for Local Hosting on Google Pixel 10 Pro

    Kamila Wojciechowska says she has found multiple clues regarding upcoming AI features that Google intends to roll out on various Pixel devices in the coming years. She references a source from Google’s gChips division for all details concerning unreleased Tensor chipsets.

    New AI Features in Development

    Wojciechowska highlights that Google is working on Video Generative ML, which is anticipated to implement AI algorithms in Google Photos and potentially YouTube Shorts, enhancing video editing capabilities. Additionally, she mentions features like ‘Speak to Tweak’ and ‘Sketch-to-Image’, the latter already seen in some recent Samsung Galaxy models.

    Innovations in Pixel Devices

    Moreover, a new ‘NanoTPU’ could be introduced with the Pixel 11 series, which would assist in monitoring sleep patterns, detecting falls, and managing sleep apnea. Google’s Stable Diffusion-based LL models are also expected to operate locally on the Pixel 10 and later, as opposed to the cloud-based system currently in use.

    Timeline for Release

    It is said that the new image signal processors (ISPs) and Tensor Processing Units (TPUs) found in the Tensor G5 and Tensor G6 are responsible for these AI-driven innovations. Reports suggest that Google is unlikely to unveil any of these features until the Pixel 10, which is expected to arrive in summer 2025. However, it’s possible that the launch could be pushed back a year to align with the rollout of Android 17 and the Pixel 11 series.

    Sources: Android Authority & Kamila Wojciechowska

  • Google Developing AI Agent to Control Web Browsers

    Google Developing AI Agent to Control Web Browsers

    According to a report from The Information, Google is developing an AI tool that can manage web browsers to make boring tasks easier, like filling out forms or reserving flights.

    Project Jarvis Unveiled

    This AI agent, known as Project Jarvis, is set to launch alongside the upcoming Gemini AI model, which might be released in December of this year. The name "Jarvis" stands for "Just Another Very Intelligent System," inspired by a fictional AI helper in the Marvel films who assists Tony Stark.

    Features of the AI Agent

    Google plans to restrict the agent’s functionality to browsers like Chrome. It will assist users with activities such as booking cinema tickets or buying goods online. People will have the ability to interact with the agent directly and give commands for various tasks.

    If this sounds a bit like something you’ve heard before, it’s because it bears a resemblance to Anthropic’s recent Claude 3.5 Sonnet, which enables app developers to "guide Claude to operate computers like humans do". OpenAI is also believed to be creating similar solutions.

    The Information, Anthropic, Reuters, Image Source.

  • Survey Reveals What Smartphone Users Want Beyond AI

    Survey Reveals What Smartphone Users Want Beyond AI

    In the past few years, top smartphone companies have really focused on adding artificial intelligence into their devices. A clear example is Apple’s advancements in AI technology. Google’s Pixel phones stand out, now featuring the Gemini AI platform. This platform offers a language model that supports natural conversations, real-time translations, text generation, and various AI-driven photography tools. Samsung and Xiaomi have also entered the competition with their own AI systems, named Galaxy AI and MiLM, respectively. Despite these innovations, it seems that many consumers are not particularly interested in AI features, indicating that they might not be a priority for most people.

    Consumer Preferences

    A large survey conducted by CNET (Computer Network) asked 2,484 Americans about what features they consider most important when buying a smartphone. The results showed that 61% preferred a big battery, while 46% looked for sufficient storage space, and 38% focused on camera quality. In contrast, only 18% of those surveyed placed importance on AI features in their purchasing decisions. Nearly half of the respondents even said they would not pay more for smartphones with AI capabilities.

    The Limits of AI Interest

    While artificial intelligence can be beneficial in daily life, the general interest seems low. One reason for this could be worries over personal data security, as 34% of participants expressed concerns in this area. Additionally, some critics claim that AI is primarily a marketing gimmick, used to create a sense of advancement without real hardware improvements. Although this critique wasn’t directly addressed in the survey, it might contribute to the overall lack of enthusiasm for AI features in smartphones.

    CNET

    Image source: Tim Dougles/Pexels