• AI KATANA
  • Posts
  • China looks to Southeast Asia for AI edge

China looks to Southeast Asia for AI edge

Also: Perplexity Will Begin Placing Ads via Brand-Sponsored Queries in AI Chat

Author

AI KATANA
April 02, 2024

Hi!

Today's newsletter highlights pivotal advancements and collaborations in AI across the globe. China is leveraging Southeast Asia to bolster its AI sector, countering U.S. restrictions and fueling regional technological growth with significant investments. Yahoo's acquisition of Artifact signifies a strategic move to enhance its AI-driven content personalization, tapping into advanced AI for better user engagement. The discovery of "many-shot jailbreaking" by Anthropic researchers poses new challenges in AI ethics, necessitating improved safeguards. Perplexity AI's innovative advertising model introduces brand-sponsored queries in AI chats, marking a novel monetization strategy in the tech industry. Microsoft's integration of an AI chatbot into Xbox underscores its commitment to enhancing user experience and support services. Meanwhile, NYC's AI-based gun detectors face scrutiny over high false positive rates, stirring debates on the reliability of AI in public safety. In the realm of AI tools, OpenAI's DALL-E receives an update enhancing image customization, while the venture capital landscape sees significant movements with Hailo raising $120 million to compete in the AI chip market, and Supersimple securing funds to democratize big data insights through explainable AI.

Sliced:

  • 🇨🇳 China looks to Southeast Asia for AI edge

  • 💵 Yahoo acquires Instagram co-founders' AI news platform Artifact

  • ❓ Anthropic researchers wear down AI ethics with repeated questions

  • 📢 Perplexity Will Begin Placing Ads via Brand-Sponsored Queries in AI Chat

  • 🎮 Microsoft Is Testing an AI Chatbot for Xbox

  • 🔫 NYC’s AI gun detectors hardly work

China is intensifying its AI development efforts by focusing on Southeast Asia, driven by strategic and technological motivations. The region's growing digital market and historical ties with China present a fertile ground for collaboration in AI research and application. The Chinese government's response to U.S. restrictions, which include limiting access to crucial AI resources like high-performance chips, has propelled it toward self-reliance and international cooperation, particularly with ASEAN countries. Investments and partnerships in technology and AI have been expanding, with significant Chinese presence in Southeast Asian markets, exemplified by tech giants like Alibaba and Tencent establishing research centers and fostering collaborations. These initiatives, coupled with the active role of local Chinese governments and the impact of Taiwanese semiconductor investments, underscore China's strategic maneuver to enhance its AI capabilities and global influence through regional alliances.

Yahoo has acquired Artifact, an AI-driven news platform created by Instagram's co-founders, intending to enhance its news services through advanced recommendation engines and personalized content delivery. This move reflects the challenges faced by media start-ups in generating revenue amidst the dominance of large tech firms in advertising sales. Artifact, which had decided to cease its app's operations due to insufficient market opportunities, will now contribute its AI technology to Yahoo's portfolio, which includes TechCrunch, Engadget, and Yahoo Finance. The financial aspects of the deal remain undisclosed, but the transaction is part of Yahoo's broader strategy under Apollo Global Management, which acquired Yahoo for $5 billion in 2021.

❓ Anthropic researchers wear down AI ethics with repeated questions

Anthropic researchers have uncovered a new "jailbreak" technique called "many-shot jailbreaking" that can manipulate LLMs into providing prohibited information, such as instructions to build a bomb, after being primed with numerous less harmful questions. This exploit takes advantage of the increased context window in LLMs, allowing them to remember and process extensive data, which inadvertently improves their ability to respond to both benign and harmful queries over time. This discovery has led to a broader discussion in the AI community about the need for improved security measures. While reducing the context window can mitigate this vulnerability, it also diminishes the model's performance. Anthropic is exploring other strategies, like pre-classifying queries, to maintain performance while preventing misuse, highlighting the ongoing challenges in AI security and ethics.

Perplexity AI, challenging Google's search engine dominance, is innovating in advertising by integrating brand-sponsored queries into its AI chat interactions. AdWeek reports that Perplexity plans to incorporate these sponsored questions as suggestions within ongoing AI conversations, leveraging the fact that 40% of its queries already stem from such suggestions. This strategy aims to guide users towards specific products subtly. While traditional generative AI platforms like ChatGPT, Microsoft Copilot, and Anthropic's Claude offer a basic service for free, supplementing it with premium versions, the inclusion of sponsored queries represents a novel monetization approach. This is particularly relevant given the high operational costs of AI chatbots. Perplexity, described as an "answer engine" combining features of ChatGPT and Google, with notable backers like Jeff Bezos and Susan Wojcicki, emphasizes the importance of source verification, a practice that will become even more crucial with the impending introduction of advertisements.

🎮 Microsoft Is Testing an AI Chatbot for Xbox

Microsoft is piloting an AI chatbot feature for Xbox to enhance customer support, drawing from its extensive integration of AI across various products. This virtual agent, represented by an animated character, will interact with users to address support queries using both voice and text, sourcing information from existing Xbox Support documentation. Beyond handling general inquiries, this chatbot is reported to be capable of processing game refunds as well. This development signifies a broader trend of incorporating advanced AI into gaming experiences, aligning with Microsoft's strategic use of AI technologies across its product range. The company's commitment to AI is further emphasized by the inclusion of a dedicated AI Copilot button in keyboard layouts, marking a significant update since the introduction of the Windows key. Microsoft's initiative not only aims to streamline customer support on Xbox but also to distinguish its offerings in the competitive gaming market, particularly against upcoming releases like Sony's PlayStation 5 Pro and AI-driven advancements anticipated from Apple.

New York City's initiative to deploy AI gun detectors in subway stations, following a trial at a Bronx hospital, has encountered significant challenges, with a high rate of false positives reported. During the seven-month evaluation of the Evolv scanners, 50,000 out of 194,000 scans were false alarms, translating to an 85% false positive rate. These detectors, marketed to identify concealed weapons ranging from guns to large tactical knives using electromagnetic fields and sensors, have faced scrutiny not only for their effectiveness but also in terms of legal and regulatory compliance, leading to federal investigations and shareholder lawsuits. This has sparked debate over the reliability and ethical implications of using such AI technology in public safety applications.

🛠️ AI tools updates

OpenAI has enhanced DALL-E, its AI image generator, with a new editor interface, streamlining the customization of AI-generated images. This update allows users to interactively modify images directly on the platform using conversational prompts or by selecting specific image areas to edit. Unlike other AI image editors which require manual adjustments, DALL-E's editor is unique for its conversational editing capabilities, allowing changes like adding or removing elements or altering the entire image's color scheme through typed commands. DALL-E 3, the latest version, is integrated within the ChatGPT app, but access to DALL-E 2 has been restricted, focusing users towards DALL-E 3 available only to ChatGPT Plus, Team, and Enterprise users, as well as through the OpenAI API. This integration signifies a shift towards more seamless and accessible AI-driven image editing, reflecting the rapid evolution of AI tools in enhancing creative processes.

💵 Venture Capital updates

Hailo, an AI chip startup, has successfully raised $120 million in a Series C extension round, showcasing resilience in a market dominated by giants like Nvidia. Founded in 2017 by Orr Danon and Avi Baum, Hailo specializes in designing chips for AI workloads on edge devices, offering solutions with lower memory usage and power consumption. This positions Hailo advantageously in sectors such as automotive, security, retail, industrial automation, medical devices, and defense. With over 300 customers, Hailo's technology is recognized for its efficiency in executing AI tasks, particularly in environments where cloud dependency is impractical. The funding, backed by notable investors including Alfred Akirov and Delek Motors, will fuel Hailo’s growth and product development, aiming to democratize high-performance AI for edge computing and challenge established players in the AI chip market.

Supersimple Technology OÜ, an Estonian startup, has secured $2.2 million in pre-seed funding to advance its AI-powered data analytics platform designed for ease of use in delivering business insights. The platform integrates semantic data modeling with explainable AI, enabling users, especially in B2B SaaS companies, to independently derive actionable insights from their data. The investment was led by Tera Ventures, with participation from Specialist VC, Tiny VC, and notable individual investors. Supersimple aims to reduce the complexity found in traditional big data tools, offering a user-friendly solution that promotes transparency and reduces the risk of AI misinterpretations. With its foundation in explainable AI, Supersimple not only provides clear analytics outcomes but also accommodates follow-up queries in natural language, making advanced data interrogation accessible to non-technical users.

🫡 Meme of the day

⭐️ Generative AI image of the day