Category: Artificial intelligence

  • OpenAI Bans China Developers, Boosting Local AI Sector Growth

    OpenAI Bans China Developers, Boosting Local AI Sector Growth

    OpenAI’s decision to restrict access for developers based in China is set to reshape the AI industry in the region. Industry experts and analysts suggest that this decision won't impede but rather boost the development of the Chinese AI sector.

    Zhou Hongyi, CEO of Qihoo 360, foresees that the restriction will steer Chinese users towards indigenous AI models. Qihoo 360 has already crafted its own large language model (LLM). Even though OpenAI’s services are officially unavailable in China, developers have been using VPNs and APIs to circumvent these restrictions. However, this new ban is eliciting a quick response from Chinese tech companies eager to seize the opportunity.

    Incentives from Local Companies

    In response to OpenAI’s ban, various Chinese firms are rolling out incentives to lure developers. Beijing-based Zhipu AI, for instance, has introduced a “special house-moving plan” to ease the transition to its platform. Prominent companies like Alibaba, Baidu, Baichuan, and 01.ai are also extending multiple perks, such as discounts, freebies, and technical support. Baidu is offering free AI model fine-tuning and 50 million free tokens, while SenseTime Group Inc. is providing 50 million free tokens and Zhipu AI is giving away 150 million tokens along with training sessions.

    source:openai.com

    Market Impact and Future Prospects

    The ban could have a significant impact on the market, potentially leading to the exit of smaller startups that emerged during the “battle of a hundred models.” There are concerns about whether other open-source models, like Meta’s Llama, will also cut off access to Chinese developers.

    This move by OpenAI is likely to benefit local LLMs by reducing competition, but Chinese developers may face challenges in accessing advanced global algorithms. This aligns with the US government’s strategy to limit Chinese access to advanced AI and semiconductor technology, affecting the broader US-China tech rivalry.

    In the long run, the lack of access to global tools might decelerate Chinese AI advancements. Alibaba Chairman Joe Tsai estimates it will take two years for Chinese AI models to reach parity with their US counterparts. This scenario might also speed up the migration of Chinese tech startups to overseas markets in search of more stable opportunities.

    Continued Access via Microsoft

    Microsoft, however, continues to provide access to OpenAI models for eligible Hong Kong customers through its Azure cloud platform, with no changes to Azure OpenAI service offerings in Hong Kong. This ensures that some developers in the region still have the tools they need.

    Overall, OpenAI’s restriction is not a setback but a catalyst for growth and transformation in the Chinese AI sector. With over 200 home-grown LLMs, including 117 approved for public release, China is well-positioned to bolster its standing in the global AI industry.

    Tags: , ,
  • Honor Launches AI Eye Protection and Deepfake Detection

    Honor Launches AI Eye Protection and Deepfake Detection

    Honor has showcased its advancements in on-device AI at MWC Shanghai 2024, with a focus on user empowerment and safety. The company introduced two AI features: AI Defocus Eye Protection and AI Deepfake Detection.

    With the rise of nearsightedness due to prolonged screen time, Honor’s AI Defocus Eye Protection offers a unique solution. By simulating defocus glasses on the device’s display, this technology induces controlled defocus in peripheral vision, slowing down the eye elongation process linked to myopia. Early tests indicate an average 13-degree reduction in transient myopia after 25 minutes of reading, with some users experiencing up to 75 degrees reduction.

    AI Deepfake Detection

    On the other hand, Honor’s AI Deepfake Detection tackles the growing threat of manipulated content and online scams. The on-device feature scrutinizes factors like eye contact, lighting consistency, image clarity, and video playback for inconsistencies often undetectable by the human eye.

    This AI is trained on a vast dataset of videos and images associated with online scams. This allows it to identify, screen, and compare content in a mere three seconds! If synthetic or altered content is detected, a user receives an immediate risk warning, safeguarding them from potential scams.

    Commitment to On-Device AI

    Honor emphasizes its commitment to on-device AI, believing it to be the key to delivering personalized services while safeguarding user privacy. Unlike cloud-based AI, which relies on remote servers, on-device AI processes data directly on the smartphone, ensuring greater control and security for users.

    Honor’s CEO George Zhao envisions a future where on-device AI empowers users, seamlessly integrating into their lives and enhancing their capabilities. With AI Defocus Eye Protection and AI Deepfake Detection, Honor is taking significant steps towards this vision, demonstrating the transformative potential of human-centric AI.

    Honor Launches AI Eye Protection and Deepfake Detection
    Tags: , ,
  • China’s Big Bet on Memory Chips: Can They Win the Global Race?

    China’s Big Bet on Memory Chips: Can They Win the Global Race?

    China faces a challenging yet promising landscape in the high-bandwidth memory (HBM) chip market and the broader semiconductor industry, driven by surging global demand and strategic expansions.

    High-bandwidth memory chips are experiencing a significant surge in demand, primarily due to their vital role in AI applications within data centers. This growing demand is expected to contribute to around an 80% revenue growth in the global memory chip market this year, recovering from a low base in the previous year. Leading this market are SK Hynix, holding a 50% global market share, followed by Samsung Electronics and the US-based Micron Technology.

    China’s Position and Challenges

    China accounts for 30-35% of global memory consumption. However, the country faces substantial challenges in producing high-end memory chips due to limitations in its semiconductor supply chain. Currently, China’s production capabilities are more suited to mid to low-end memory solutions. Consequently, as China’s AI ecosystem continues to expand, it will increasingly rely on imports, particularly from Korean memory chip producers.

    Despite these challenges, China is actively working to strengthen its position in the HBM market. ChangXin Memory Technologies (CXMT) represents China’s leading hope for domestic HBM production. CXMT, in collaboration with TongFu Microelectronics, is developing HBM samples, although it may take up to four years for these products to reach the market. In the interim, China sees potential growth opportunities in lower-end memory products, particularly for AI edge devices such as autonomous vehicles, AI-enabled smartphones, and personal computers. By 2025, the demand from edge AI applications is expected to drive about 20% industry revenue growth.

    External Challenges

    China’s ambitions in the semiconductor sector are also constrained by external factors, including potential further restrictions from the US, Japan, and the Netherlands on developing HBM chips.

    Strategic Expansion Amid US Sanctions

    In response to fears of more US sanctions, Chinese semiconductor companies, including SMIC and Hua Hong Semiconductor Group, are ramping up their capacities. Investments are primarily focused on legacy chips used in applications like cars and consumer electronics. This year, China’s wafer fabrication capacity is projected to increase by 15% to 8.9 million wafers per month, with a further 14% increase to 10.1 million wafers per month expected next year. This rapid growth is set to make China account for about 30% of the world’s total wafer fabrication capacity, outpacing global growth rates.

    Economic and Market Impact

    The expansion efforts have led to a significant surge in sales of semiconductor wafer fab equipment in China, which saw a 48% increase last year compared to a mere 1% worldwide growth rate. Meanwhile, China’s imports of integrated circuits (ICs) dropped by 10.8% in volume and 15.4% in value.

    Despite these aggressive expansions, analysts warn of potential overcapacity in the next two years, which could drive global chip prices down. Additionally, the Biden administration’s planned tariffs on $18 billion worth of Chinese imports, including a 50% hike on semiconductor imports, add to the complexities.

    However, China’s drive towards self-sufficiency has benefited local foundries like SMIC and Yangtze Memory Technologies Corp. These foundries enjoy higher capacity utilization rates due to domestic substitution policies. For example, Hua Hong Semiconductor operates at maximum capacity and plans a 10% price increase in the latter half of the year.

    Local foundries in China are showing a faster recovery in capacity utilization compared to their global peers, supported by high customer demand during the traditional peak inventory stocking season. Recent price adjustments aim to alleviate profit pressures rather than indicating a full recovery in demand.

    China is strategically expanding its semiconductor capacity to mitigate the impact of US sanctions and reduce reliance on imports. While rapid growth brings the risk of overcapacity, domestic policies, and self-sufficiency drives provide significant advantages, positioning China as a formidable player in the global semiconductor market.

    Tags: , ,
  • Android Circle to Search May Soon Include Audio and Music Features

    Android Circle to Search May Soon Include Audio and Music Features

    Samsung's Galaxy S24 series has introduced Galaxy AI, a suite of artificial intelligence-powered features designed to simplify everyday tasks. Among the standout features of Galaxy AI is Google's Circle to Search, which is now accessible on Pixel devices as well.

    The Circle to Search feature allows users to obtain information about anything on their screen by simply drawing a circle around it. Recently, new reports indicate that Google is working on enhancing Circle to Search by adding the capability to recognize songs and popular audio clips.

    New Audio Search Button

    The information comes from @AssembleDebug, who uncovered it during an APK teardown, as noted by Android Authority. According to the report, the latest beta version of the Google app, which supports Circle to Search on Android devices, includes a code string that references a new Audio search button.

    <string name="omnient_zerostate_audio_search_button_content_description">Audio search button</string>

    AssembleDebug succeeded in enabling the UI for this new button, which resembles a musical note. The button appears alongside the existing button for on-screen text translation.

    Functionality and Speculations

    Currently, the button is non-functional, and Google has not provided an official explanation regarding its purpose. However, based on the musical note icon, it is speculated that clicking the button could allow Circle to Search to use the phone's microphone to analyze ambient sounds.

    If a song or well-known audio clip is detected, the feature would then offer relevant information. There is also the possibility that Circle to Search might allow users to hum a tune for identification. It is worth noting that there is already a microphone icon in the search bar that offers similar song search capabilities.

    If this feature operates as anticipated, it would make song identification and information retrieval simpler for Android users. Hopefully, the upcoming beta version of the Google app will shed light on the button's functionalities by activating them. We will keep you updated with any new details about the feature as they become available.

    Tags: , ,
  • Amazon’s Alexa Adds Monthly Subscription Fee for Prime Members

    Amazon’s Alexa Adds Monthly Subscription Fee for Prime Members

    Amazon is preparing to revamp Alexa, though accessing its most advanced features may come with a price tag. Reports indicate a potential $5 to $10 monthly subscription fee in addition to a Prime membership to unlock the AI-enhanced Alexa. This upgraded assistant is slated for release in August 2024 and promises several enhancements.

    Enhanced Capabilities

    The new advanced Alexa is expected to offer personalized advice for activities like shopping or art. The need to say “Alexa” repeatedly might be eliminated. The updated version could process and complete multiple requests simultaneously, such as drafting an email and ordering takeout at the same time. It might even adapt to user habits, customizing routines like starting the coffee maker when the alarm sounds. Additionally, this sophisticated Alexa is anticipated to provide nuanced shopping recommendations.

    Free Tier Still Available

    Despite these changes, Amazon will not completely phase out the free tier. A basic version of Alexa with new generative AI features will remain accessible. This strategy appears to be a response to the escalating competition in the AI assistant market, with rivals like Google, Microsoft, OpenAI, and Apple's improved Siri all vying for user engagement. Alexa’s future strategy seems to focus on a dual approach: maintaining a free basic service while introducing a premium tier with advanced functionalities.

    Tags: , ,
  • Pixel 9 May Launch with Exclusive “Creative Assistant”

    Pixel 9 May Launch with Exclusive “Creative Assistant”

    Android 15 Beta 3 may have unveiled a new app under development for Google’s Pixel series: Creative Assistant. Unearthed by Mishaal Rahman from Android Authority, the app’s package name, com.google.android.apps.pixel.creativeassistant, suggests it could be a feature unique to Pixel devices.

    According to the findings, Creative Assistant is an innovative tool that employs AI to produce stickers and potentially emojis, akin to Apple’s recently introduced Genmoji.

    The app itself isn’t accessible in the new Android Beta, but references to it were found within the Markup app included in the Beta. For those who might not know, Markup is an application for editing screenshots and adding annotations in apps such as Google Photos.

    Features and Integration

    The code in the Markup app indicates a “remix” button connected to Creative Assistant, enabling users to create AI-generated stickers and incorporate them into images, as per Mishaal’s analysis.

    There’s a possibility that the Creative Assistant feature may operate on-device rather than relying on cloud processing. The anticipated Tensor G4 chip in the Pixel 9, along with Google’s efficient AI model Gemini Nano, could manage sticker generation, given their smaller data size compared to full images.

    Comparison with Existing Tools

    The feature seems reminiscent of Google’s current AI-driven creative tools. For instance, Gboard’s Emoji Kitchen lets users merge existing emojis to craft new ones, while Creative Assistant will leverage Generative AI to enable the creation of entirely new emojis.

    However, it remains uncertain if it can rival the versatility of Apple’s Genmoji, which allows users to design personalized emojis directly within their messaging applications.

    Future Availability

    More details are needed to compare Creative Assistant with Genmoji and assess its accessibility. Will it be a Pixel-exclusive feature, or will it eventually be rolled out to a broader Android audience? The complete details are yet to be disclosed.

    Tags: , ,
  • Samsung Integrates AI with Smart Home Products

    Samsung Integrates AI with Smart Home Products

    Samsung is extending its AI capabilities beyond smartphones into smart home products. According to Businesskorea, the company may launch AI-integrated home appliances as soon as next year. This move aims to establish a "super-connected ecosystem," enhancing Samsung's competitive edge against rivals like Apple and Google.

    On-Device Processing for Enhanced Privacy

    The appliances set for a 2025 release are expected to run large language models (LLMs) directly on the devices, which benefits user privacy. Samsung plans to integrate Bixby voice assistant controls into its smart home products by July this year. Initially, the company will use cloud-based solutions to run the LLMs.

    First AI-Enabled Products

    Samsung's first AI-enabled smart home products will include Family Hub refrigerators, washing machines, and induction cookers from the Bespoke AI line, which features LCD screens. These appliances will possess AI capabilities, such as translation, that are currently limited to smartphones.

    Enhanced by the latest large language models, Bixby voice assistant will enable smart home products to understand more intricate commands and offer a more natural user experience. Additionally, these products will reportedly remember past conversations, providing further benefits.

    The number of smart home products connected to Samsung SmartThings has already exceeded 20 million, doubling from around 10 million in 2022. This figure is expected to climb to 30 million by next year.

    Tags: , ,
  • Google DeepMind AI Creates Music and Sound for Silent Videos

    Google DeepMind AI Creates Music and Sound for Silent Videos

    Google's DeepMind has unveiled a new AI tool capable of generating background music and sound effects for silent videos. This "video-to-audio" system aims to simplify the video editing process, especially for content creators.

    Currently under development, this technology offers some intriguing capabilities. Here’s an overview of the process:

    User Input

    Creators start by uploading their silent video and can include keywords or phrases to guide the AI in producing the appropriate soundscape. For instance, a silent video featuring someone walking in the dark might benefit from prompts such as “movies, horror films, music, tension, footsteps on concrete” to help the AI grasp the mood and context.

    AI in Action

    DeepMind’s AI model begins by breaking down the video to analyze its visuals. This visual data is then paired with the user-provided text prompts. Through a diffusion model, the AI processes this combined information iteratively, eventually creating background sounds that match the video content.

    Tailoring the Soundscape

    The model can generate different audio options for a single video, allowing creators to select the best match for their project. DeepMind’s system can also take into account the emotional tone of the prompt words. For example, prompts that emphasize “tension” might produce suspenseful background music, whereas prompts like “joyful celebration” could result in more upbeat sounds.

    Looking forward, DeepMind is continuously refining this technology. Future plans include enabling the AI to generate sounds automatically based solely on the video content, eliminating the need for user prompts. Additionally, they aim to enhance the system’s ability to synchronize generated dialogue with the characters’ lip movements in the video.

    This "video-to-audio" technology has the potential to transform video editing, particularly for creators who do not have access to professional audio tools or expertise.

    Tags: , ,
  • Nvidia Surpasses Microsoft as World’s Most Valuable Company

    Nvidia Surpasses Microsoft as World’s Most Valuable Company

    Nvidia (NVDA) has overtaken Microsoft (MSFT) to become the most valuable company in the world by market capitalization. This remarkable rise is driven by Nvidia’s leadership in the artificial intelligence (AI) hardware sector. Nvidia’s stock price surged 3.5% to $135.58, bringing its market cap to an astonishing $3.335 trillion. This move edged out Microsoft’s $3.317 trillion market cap and Apple’s $3.286 trillion valuation. Over the past year, Nvidia’s stock has nearly tripled, highlighting the soaring demand for its AI processors.

    Rapid AI Growth

    The past few years have seen rapid AI growth, significantly contributing to Nvidia’s expansion. This upward trend mirrors widespread investor optimism about the future of AI technology. Nonetheless, some analysts caution that this enthusiasm might diminish if there are indications of a slowdown in AI technology spending.

    A key driver of Nvidia’s success is its dominance in the AI hardware market. Their processors are renowned for being superior to those of their competitors, resulting in a situation where demand far exceeds supply. Consequently, Nvidia has become the most traded company on Wall Street, with an average daily turnover of a staggering $50 billion.

    Record-High Stock Price

    Recently, Nvidia’s stock price reached an all-time high, adding over $110 billion to its market cap—equivalent to the entire value of Lockheed Martin. The company’s market value has expanded dramatically, escalating from $1 trillion to $2 trillion in just nine months, and hitting $3 trillion in a little over three months.

    Nvidia has consistently surpassed Wall Street’s revenue and profit expectations, driven by the booming demand for its graphics processors as businesses rush to integrate AI applications. The company recently executed a 10-for-one stock split, potentially making its shares more appealing to individual investors. Despite the stock’s appreciable gains, the company’s earnings valuation has actually decreased due to rising analyst predictions for future earnings.

    Tags: , ,
  • Apple Intelligence Features Delayed to Next Year: Here’s Why

    Apple Intelligence Features Delayed to Next Year: Here’s Why

    Apple introduced new AI capabilities branded as Apple Intelligence at the annual WWDC event last week. These features are set to be incorporated into iOS 18, iPadOS 18, and macOS Sequoia for iPhones, iPads, and Macs, respectively. Although the new software versions will become available to the public later this year, they won’t come preloaded with all Apple Intelligence features. According to Bloomberg’s Mark Gurman, the rollout will be gradual, extending into 2025.

    Apple Intelligence’s Gradual Rollout

    Gurman reports that Apple plans to introduce the new Intelligence features in phases over several months. The Cupertino-based tech giant aims to avoid the issues that have plagued other AI systems. Typically, Apple releases features announced at WWDC in September when the new iPhones are launched. However, this time, Apple has disclosed that the latest AI technology will not adhere to a strict schedule. During the recent keynote, the company outlined a roadmap for what will be available later this year and early next year.

    Strategic Delays and Resource Management

    Gurman points out that the delayed rollout of Apple Intelligence has its advantages. It allows Apple to manage its resources more effectively by not overwhelming its staff with the simultaneous release of all features. Initially, Apple will focus on American English, giving the company more time to train its AI models in other languages, a process that could take years. Additionally, the tech giant will take its time to bolster its cloud infrastructure to prevent server crashes due to high user demand. Apple also aims to improve the accuracy of information delivered by its AI services by starting with a smaller user base.

    Some features of Apple Intelligence, including the enhanced Siri assistant capabilities showcased at WWDC—such as contextual searches and on-screen awareness—will not be available until next year. The ChatGPT integration is also slated for a 2025 release. On the plus side, the initial version of Apple Intelligence will include features like notification prioritization, webpage summarization, writing tools, image generation, and Genmoji.

    Non-AI Features

    Gurman also mentions that non-AI features, such as updates to the Mail app and the ability to edit spatial videos, will not be included in the initial software version updates.

    Tags: ,