Category: Artificial intelligence

  • Microsoft Phi-3-Vision Model Enhances Mobile Image Analysis

    Microsoft Phi-3-Vision Model Enhances Mobile Image Analysis

    Microsoft is broadening its Phi-3 series of small language models with the launch of Phi-3-vision. Unlike its counterparts, Phi-3-vision isn’t limited to text processing — it’s a multimodal model capable of analyzing and interpreting images as well.

    The model excels at object recognition in images

    This 4.2 billion parameter model is optimized for mobile devices and excels at general visual reasoning tasks. Users can pose questions to Phi-3-vision about images or charts, and it will provide insightful answers. While it isn’t an image generation tool like DALL-E or Stable Diffusion, Phi-3-vision is exceptional at image analysis and comprehension.

    Expansion of the Phi-3 family

    The introduction of Phi-3-vision follows the release of Phi-3-mini, the smallest model in the Phi-3 family with 3.8 billion parameters. The complete family now consists of Phi-3-mini, Phi-3-vision, Phi-3-small (7 billion parameters), and Phi-3-medium (14 billion parameters).

    Emphasis on smaller models

    This emphasis on smaller models highlights a growing trend in AI development. Smaller models require less processing power and memory, making them perfect for mobile devices and other resource-constrained settings. Microsoft has already achieved success with this strategy, as its Orca-Math model has reportedly outperformed larger competitors in solving math problems. Phi-3-vision is currently available in preview, while the rest of the Phi-3 series (mini, small, and medium) can be accessed through Azure’s model library.

    Tags: , ,
  • iOS 18: Notification Summaries & AI Photo Editing Report

    iOS 18: Notification Summaries & AI Photo Editing Report

    Apple’s annual Worldwide Developer Conference (WWDC) is scheduled for June 10. As expected, the tech giant from Cupertino will unveil the next-generation software for its products, including iPhones and iPads. The primary highlight of the event is anticipated to be iOS 18, which is rumored to incorporate AI features. In his latest update, Bloomberg’s Mark Gurman has shed light on the AI functionalities that the new operating system might introduce for iPhones.

    iOS 18 to Feature On-Device AI Capabilities

    According to Gurman’s newsletter, iOS 18 will include a notification summarization tool that can condense notifications, news articles, and transcribe voice memos. Siri is also expected to receive enhancements for a more conversational tone. As a result, the upcoming software version will emphasize proactive intelligence to assist users in their daily lives. Additionally, the report suggests that Apple might introduce AI-based photo editing tools and improvements to the Calendar app.

    Gurman further notes that Apple will rely on on-device processing for these AI features. The company is also contemplating offering AI services via the cloud, supported by Apple silicon chips in its data centers. However, the tech giant will not be announcing its proprietary chatbot at this time, as it is currently behind in the Gen AI space. Gurman hints that Apple could reveal a partnership with OpenAI at the WWDC, with the possibility of launching a deeply integrated chatbot later on.

    Potential Partnership with Google for Gemini AI

    Apple has also been in discussions with Google regarding the integration of Gemini AI into iOS 18, although no agreement has been finalized yet. Nonetheless, the company is poised to make its entry into the AI arena in the coming weeks. In addition to today’s report on AI features, previous reports indicated that iOS 18 might revamp some native apps and introduce changes to the home screen.

    Tags: ,
  • ChatGPT Update: Analyze Excel Sheets, Import from Google Drive & OneDrive

    ChatGPT Update: Analyze Excel Sheets, Import from Google Drive & OneDrive

    OpenAI has elevated ChatGPT's data analysis features, simplifying data exploration and manipulation significantly. This enhancement optimizes workflows and provides instant data insights.

    Seamless Integration with Cloud Storage

    One of the most notable improvements is integration with cloud storage services. The need to download and upload files manually is now obsolete. ChatGPT allows direct access to data stored in Google Drive and Microsoft OneDrive, streamlining the entire process and enabling seamless work with cloud-stored documents.

    FILE PHOTO: OpenAI and ChatGPT logos are seen in this illustration taken, February 3, 2023. REUTERS/Dado Ruvic/Illustration/File Photo

    Enhanced Data Visualization

    Data visualization has received a substantial upgrade. ChatGPT now supports interactive table and chart views, letting users explore their data in real-time. These interactive visualizations offer a more intuitive grasp of the data, facilitating deeper insights.

    Customization of charts is another key feature. ChatGPT allows you to tailor the charts to meet your specific requirements, aligning them with your presentation or report. Once customized, these charts can be easily downloaded for seamless inclusion in your work.

    Advanced Data Handling

    The capabilities for data handling have also been substantially improved. Whether dealing with large datasets, cleaning messy data, or generating insightful charts, ChatGPT can manage these tasks effortlessly. This enhanced capability is powered by a new underlying model, enabling ChatGPT to tackle even the most complex data tasks with ease.

    Upcoming Innovations

    Concurrently, discussions about a new model named ADA V2, speculated to be GPT-4, have surfaced with the ChatGPT update. Users involved in the grayscale testing of this model have praised its robust coding features.

    OpenAI's rapid pace of innovation is evident. Just days after the GPT-4o reveal, significant advancements with both ChatGPT and ADA V2 are apparent. These innovations are swiftly transforming the data analysis field, with the potential for an even more powerful "GPT-5" on the horizon, sparking excitement.

    Tags: ,
  • Microsoft Proposes Relocation for China AI Staff amid US-China Tensions

    Microsoft Proposes Relocation for China AI Staff amid US-China Tensions

    Microsoft has revealed a noteworthy relocation offer for its AI staff based in China. The company is providing these employees the option to move to countries such as the United States, Australia, and Ireland. This decision affects approximately 700 to 800 employees, primarily those working on machine learning in the Azure cloud computing division. A few of these employees might also have opportunities for international rotations.

    Relocation Decision Deadline

    Employees need to decide by June 7 whether to relocate or continue in their current roles within China. This initiative comes as Microsoft pauses new hiring in China, impacting its offices in Beijing, Shanghai, and Suzhou. Nevertheless, Microsoft reassures its continued commitment to its operations in China and other international markets.

    Geopolitical Context

    This relocation offer mirrors broader geopolitical issues, especially the intensifying US-China tech rivalry. AI technology has become a significant point of contention. The Biden administration is contemplating new restrictions on exporting proprietary AI models to China, adding to the existing limitations on Chinese firms’ access to advanced semiconductors and chip-making tools. Microsoft is navigating these tensions while continuing to pursue business for its AI services in mainland China and Hong Kong.

    Strategic Relocation

    Last year, Microsoft transferred some top AI researchers from China to a new lab in Vancouver, Canada. This lab is part of a global initiative to integrate talent from various countries, including China. The current relocation offer is seen as a strategic response to the ongoing trade and tech disputes between the US and China.

    The US has recently increased tariffs on several Chinese imports, including electric vehicles and semiconductors, further straining relations. In response, China has vowed to take measures to safeguard its interests. Despite these challenges, Microsoft’s long-term presence in China, dating back to 1992, highlights its commitment to maintaining operations in the region.

    As the tech industry adjusts to geopolitical changes, companies like Microsoft are making strategic decisions to ensure their operations and talent pools remain strong. The relocation offer to China-based AI employees is a part of these efforts, reflecting the intricate interplay of global business and international relations.

    Tags: , ,
  • Baidu’s Wenxin AI Gains Traction with Xiaomi, Lenovo, Vivo & NIO

    Baidu’s Wenxin AI Gains Traction with Xiaomi, Lenovo, Vivo & NIO

    Baidu, the Chinese technology behemoth, is making notable advancements in artificial intelligence (AI) through its Wenxin Big Model. The company’s latest financial report showcased positive financial outcomes and an increase in the adoption of its flagship AI product.

    Cost Efficiency Drives Growth: Baidu’s Affordable Wenxin Model Accelerates Adoption

    Despite modest year-on-year revenue growth, Baidu’s net profit saw a robust 22% rise. This improvement was partly due to the growing acceptance of Wenxin. Initially integrated with smartphones from China Samsung and Honor, Wenxin has now formed partnerships with major brands like Xiaomi, OPPO, and Vivo.

    This development marks a significant leap for Baidu’s AI goals. Wenxin is expanding beyond smartphones, entering the personal computer (PC) market through a collaboration with Lenovo. Additionally, the electric vehicle (EV) sector is showing interest, with NIO becoming one of Wenxin’s partners.

    Li Yanhong, Baidu’s CEO and co-founder, believes that integrating Wenxin with smart devices paves the way for widespread adoption among a broader audience. This strategic initiative positions Baidu as a pivotal player in the rapidly growing AI infrastructure sector.

    Wenxin One Word 3.5: A Leap in Cost Efficiency

    Moreover, Baidu is focused on reducing Wenxin’s inference cost. The latest version, Wenxin One Word 3.5, offers an impressive 99% reduction in inference cost compared to version 3.0. This substantial decrease makes Wenxin more appealing to businesses exploring and creating AI-powered applications based on the Wenxin One Word platform.

    Li Yanhong underscores the transformative impact of generative AI in China. He foresees foundational models like Wenxin becoming a crucial part of essential infrastructure, seamlessly integrated into various aspects of daily life. Baidu’s dedication to affordability and efficiency with the Wenxin Big Model series is a strategic approach likely to unlock new opportunities for the company.

    With ongoing advancements and strategic collaborations, Baidu’s Wenxin Big Model is set to become a significant player in the Chinese AI arena. As the generative AI era progresses, Wenxin’s integration across diverse tech sectors has the potential to revolutionize how we interact with technology and navigate the digital world, providing a compelling alternative to existing solutions like ChatGPT.

    Tags: , ,
  • Xiaomi’s MiLM LLM Approved for Smartphones, Cars, and More Devices

    Xiaomi’s MiLM LLM Approved for Smartphones, Cars, and More Devices

    Xiaomi’s large language model (LLM), known as MiLM, has successfully completed the registration process for large models, as announced on the company’s Weibo account.

    With this milestone, Xiaomi indicates that MiLM is prepared for incorporation into its range of products, such as smartphones, smart home devices, and even Xiaomi automobiles. The announcement also hinted at the potential of expanding MiLM’s capabilities to a broader audience in the future.

    Benchmark Achievements

    MiLM made its public debut in August 2023 on benchmark platforms C-Eval and CMMLU, where it delivered impressive performance.

    The model secured the top position within its parameter category on the C-Eval leaderboard and ranked 10th overall. According to the project’s GitHub page, MiLM-6B, the specific variant in question, boasts 6.4 billion parameters.

    Subject-Specific Performance

    C-Eval’s subject-specific breakdown showcases MiLM-6B’s proficiency in STEM fields (Science, Technology, Engineering, and Mathematics). The model achieved high accuracy scores across all 20 STEM subjects, including metrology, physics, chemistry, and biology.

    While MiLM-6B shows strong performance in most liberal arts subjects, areas requiring “abstract thinking” like law, mathematics, programming, and probability theory appear to need further development.

    Social Sciences and Humanities

    In the realm of social sciences, MiLM-6B demonstrated good accuracy in eight out of ten subjects, with education and geography being the exceptions. As for the humanities, the model performs admirably in history and law, though the accuracy in other subjects is yet to be fully assessed.

    With MiLM-6B overcoming significant hurdles, it’s now set to be integrated into various Xiaomi products. Despite its varied performance across different subjects, it shows promise for enhancing user experiences in a wide range of applications.

    Tags: , ,
  • Sundar Pichai Addresses OpenAI’s Alleged Unauthorized YouTube Use

    Sundar Pichai Addresses OpenAI’s Alleged Unauthorized YouTube Use

    OpenAI has introduced an impressive text-to-video tool named Sora, capable of generating lifelike video clips from simple text prompts. Since the release of this tool, there has been ongoing curiosity about the data used to train the model.

    Training Data Controversy

    When asked in an interview if YouTube videos were used to train the model, OpenAI's CTO couldn't provide a definite answer, saying, "I’m not sure about it." Similarly, the COO declined to confirm whether the model was trained using YouTube content. Despite these ambiguous responses, reports have surfaced alleging that OpenAI utilized YouTube videos for training Sora.

    In recent developments, Google’s CEO Sundar Pichai addressed the issue, stating that he would resolve it if the allegations prove to be accurate. According to a New York Times article, OpenAI employed over a million hours of YouTube content for Sora's training.

    Google's Response

    When questioned about potential violations of Google’s terms and conditions, Sundar Pichai responded, "Look, I think it’s a question for them to answer. I don’t have anything to add. We do have clear terms of service." He further mentioned, "And so, you know, I think normally in these things we engage with companies and make sure they understand our terms of service. And we’ll sort it out."

    Reportedly, The New York Times has already taken legal action against OpenAI for using their copyrighted content in AI training. However, Pichai did not disclose his strategy for addressing this issue.

    Creator Rights and AI Training

    Ideally, content creators should have the right to opt in or out of having their material used by others. AI training necessitates a vast amount of data, typically sourced from the internet, but this should be done with proper permission. When asked if YouTube content was used by OpenAI, the company’s COO hinted at future plans. He mentioned that alongside developing a tool to detect AI-generated images, they are also working on a "content ID system for AI" that would allow creators to see where their content is being used, who is training on it, and to opt in or out of such training.

    Tags: , ,
  • Google Unveils 6th Gen TPU with 4.7x More Computing Power

    Google Unveils 6th Gen TPU with 4.7x More Computing Power

    Google introduced the sixth generation of its Tensor Processing Unit (TPU) for data centers, named Trillium, at the I/O 2024 Developer Conference today. Although a specific release date wasn't mentioned, Google confirmed that Trillium would be available later this year.

    Enhanced Memory Bandwidth and Performance Gains

    Google CEO Sundar Pichai highlighted the company's continuous commitment to AI advancements, stating, “Google was born for this moment. We have been a pioneer in GPUs for more than a decade.”

    Pichai then showcased the remarkable performance enhancements of Trillium. Compared to the fifth generation TPU, Trillium offers an astounding 4.7 times increase in computing power per chip. This leap was made possible by improving the chip’s matrix multiplication unit (MXU) and increasing the overall clock speed. Furthermore, Trillium benefits from doubled memory bandwidth.

    Third-Generation SparseCore Technology

    Trillium integrates Google’s third-generation SparseCore technology, described as “a purpose-built accelerator for common large-scale tasks in advanced ranking and recommendation workloads.” This advancement enables Trillium TPUs to train models more swiftly and provide lower latency when serving those models.

    Focus on Energy Efficiency

    Energy efficiency was another major focus for Google. Pichai emphasized Trillium as the company’s “most energy-efficient” TPU to date. This is especially important given the increasing demand for AI chips, which can significantly impact the environment. Google claims that Trillium delivers a 67% improvement in energy efficiency compared to the previous generation.

    Tags: , ,
  • OpenAI Releases GPT-4o: Enjoy GPT-4 Premium Features for Free

    OpenAI Releases GPT-4o: Enjoy GPT-4 Premium Features for Free

    OpenAI has introduced a new model, GPT-4o, which will become available to the public over the coming weeks. This new model incorporates premium features of GPT-4 and includes an updated web user interface. During the launch event, OpenAI’s CTO Mira Murati showcased several capabilities of this advanced model. Let's explore them in detail.

    GPT-4o Announcement

    GPT-4o is designed to be more efficient, with enhanced abilities to process both auditory and visual inputs. OpenAI describes this as "a step towards much more natural human-computer interaction." The model can now handle text, images, and audio input, offering seamless assistance to its users. The voice mode has been significantly improved, providing quicker responses and better comprehension.

    Previously, the voice mode required three separate models for transcription, intelligence, and text-to-speech functions, which often resulted in delays. In contrast, GPT-4o integrates these functions natively, enabling smoother performance. Using your phone's camera, you can easily share information with the model and ask questions using the voice mode. The new model can respond to voice inputs in just 232 milliseconds, closely matching human response times. It also offers responses in various tones to suit user preferences and has better and faster comprehension of non-English languages compared to GPT-4 Turbo. Additionally, GPT-4o can function as an interpreter.

    API and Premium Features

    GPT-4o will also be accessible via API, allowing developers to build and enhance AI applications using its advanced capabilities. While the new model's features are available for free, premium users will have access to five times the resources compared to the standard offering.

    OpenAI has also released a ChatGPT app for macOS-based desktops. This app provides deeper integration into the macOS platform, aiming to simplify user workflows. With a keyboard shortcut (Option + Space), users can quickly access the tool's conversation page.

    In summary, GPT-4o brings several improvements and new features, enhancing the efficiency and versatility of human-computer interactions. The new model's capabilities, combined with the new app for macOS, aim to offer a more integrated and seamless user experience.

    Tags: ,
  • SoftBank-backed Arm to Launch AI Chip in 2025

    SoftBank-backed Arm to Launch AI Chip in 2025

    There is a new player in the realm of Artificial Intelligence with Arm Holdings, a part of the SoftBank Group, stepping into AI chip development. The initiative aligns with SoftBank CEO Masayoshi Son's grand plan to invest $64 billion to establish the conglomerate as a frontrunner in artificial intelligence.

    Arm, a prominent UK-based company known for its chip designs, is gearing up to introduce its initial AI chip products by 2025. To jumpstart this effort, Arm will create a specialized AI chip division, with intentions to reveal a prototype by early 2025. Production will kick off in the autumn of the same year, overseen by contract manufacturers.

    Arm's Foray into AI Chips

    Funding for this venture will be shared by Arm and SoftBank, with discussions ongoing with major semiconductor manufacturers like Taiwan Semiconductor Manufacturing Corp (TSMC) to secure production capabilities.

    Looking towards the future, there are suggestions that once the mass production operations are established, Arm's AI chip business might be spun off and integrated within the SoftBank ecosystem.

    SoftBank's Diversification Strategy

    Arm's strategic maneuver comes amid SoftBank's broader efforts to diversify its investments and decrease reliance on dominant players such as Nvidia. CEO Masayoshi Son envisions leveraging AI, semiconductor, and robotics technologies to transform multiple industries, fostering innovation and expansion.

    The market outlook for AI chips appears promising, with analysts projecting substantial growth, potentially exceeding $200 billion by 2032. SoftBank views this as a prime opportunity to capitalize on rising demand and bypass the constraints imposed by existing market players.

    SoftBank's Financial Trajectory

    Financially, SoftBank is on a recovery path, aiming to rebound from prior setbacks. With substantial cash reserves at hand, the conglomerate is well-equipped to support its ambitious investment strategies across diverse sectors, including AI, data centers, and renewable energy.

    Nevertheless, this endeavor is not devoid of risks. SoftBank has a history of adapting to technological changes, but substantial investments always entail uncertainties, testing the resilience of SoftBank's strategic foresight.

    Tags: , ,