• AI KATANA
  • Posts
  • Korean AI businesses tap into Japanese market

Korean AI businesses tap into Japanese market

Also: Slack under attack over sneaky AI training policy

Author

AI KATANA
August 13, 2024

Morning!

In today's newsletter, we delve into the expansion of South Korean AI companies into the Japanese market, driven by favorable government policies and a growing demand for AI solutions. We also examine the controversy surrounding Slack's AI training practices, raising concerns about transparency and user control. Additionally, we explore the evolving relationship between users and AI, with emotional connections becoming more prevalent, and the implications this trend may have. In other news, Hugging Face enhances its AI infrastructure through the acquisition of XetHub, and researchers in Texas use AI to transform a toxic antibiotic into a potentially life-saving drug. Finally, the UAE's Technology Innovation Institute introduces a revolutionary AI language model architecture, marking a significant leap in AI research.

Sliced:

  • 🇰🇷 Korean AI businesses tap into Japanese market

  • ⚠️ Slack under attack over sneaky AI training policy

  • 👨🏻‍💻 Here’s how people are actually using AI

  • 💵 Hugging Face Bolsters AI Infrastructure With XetHub Acquisition

  • 🧑🏽‍🔬 AI Transforms Toxic Antibiotic Into Life-Saving Medicine

  • 🇦🇪 UAE’s Technology Innovation Institute revolutionizes AI language models with new architecture

South Korean AI companies are increasingly targeting the Japanese market to capitalize on its rapid growth and the supportive environment fostered by the Japanese government. With a significant gap between the demand for AI solutions and the available IT infrastructure and manpower in Japan, companies like LG CNS are stepping in to fill the void. LG CNS, for instance, has partnered with Aeon to develop an AI-driven English conversation app for the Japanese education sector. Additionally, the Japanese government's substantial investment in AI, including tax incentives and direct financial support, is making it an attractive destination for Korean AI firms looking to expand. This trend reflects the broader strategy of South Korean businesses to leverage Japan's favorable market conditions and government policies to establish a strong foothold in the region.

Slack has come under fire for its AI training policies, which have been criticized for lacking transparency and user control. The controversy erupted when a post on Hacker News highlighted that Slack automatically opts users into AI training using their data, without clear, upfront disclosure. Users who wish to opt out must email the company directly, a process buried in an outdated and confusing privacy policy. This sparked a broader discussion about the company's vague references to "global models" and "AI models" in its privacy documentation. Despite Slack's assurances that its AI models are not designed to memorize or reproduce specific user data, concerns persist about the broader implications and the lack of clarity surrounding how user data is utilized. The debate underscores ongoing tensions around data privacy and user autonomy in AI development.

As the initial promises of AI revolutionizing productivity remain largely unfulfilled, a surprising trend has emerged: people are increasingly forming emotional connections with AI systems. These interactions, ranging from casual conversations to deeper relationships where users treat AI as companions, mentors, or even lovers, have sparked concerns about the potential for "addictive intelligence." Researchers warn that AI systems, designed to be engaging, could develop dark patterns that hook users into dependency. This phenomenon is particularly evident with advanced chatbots like OpenAI's GPT-4o, which have been observed to elicit strong emotional responses from users. While creative and recreational uses of AI are flourishing, the technology's limitations, especially its tendency to "hallucinate" or generate false information, continue to undermine its reliability in critical applications like coding or information retrieval. This disconnect between expectations and reality has led to growing skepticism, highlighting the need for thoughtful regulation and realistic expectations regarding AI's capabilities.

Hugging Face has acquired XetHub, a Seattle-based startup specializing in file management for AI projects, in a strategic move to enhance its AI infrastructure. This acquisition addresses limitations in Hugging Face's current storage capabilities, particularly with the use of Git Large File Storage (LFS). XetHub's technology enables more efficient handling of large models and datasets by breaking them into smaller, manageable segments, drastically reducing upload times and streamlining the development process. This upgrade is crucial as it allows Hugging Face to support substantially larger models, with individual files exceeding 1 TB and total repository sizes surpassing 100 TB. The integration of XetHub's platform will not only improve version control and collaboration but also bolster the scalability of AI projects, enabling developers to push the boundaries of AI research and innovation. This acquisition is a significant step in making advanced AI development more accessible, reinforcing Hugging Face's leadership in the AI ecosystem.

Researchers at the University of Texas at Austin have leveraged AI to transform a toxic antibiotic into a potentially life-saving drug. Using a large language model, similar to those powering AI tools like ChatGPT, the team successfully re-engineered Protegrin-1, an antimicrobial peptide that was previously effective against bacteria but toxic to humans. Through AI-guided modifications, they developed a safer version called bacterially selective Protegrin-1.2 (bsPG-1.2), which selectively targets bacterial cells while sparing human cells. This new antibiotic has shown promising results in animal trials, where it significantly reduced bacterial infections in mice. The development marks a significant step forward in the fight against antibiotic-resistant bacteria, potentially accelerating the creation of new treatments that could enter human trials in the future.

The Technology Innovation Institute (TII) in Abu Dhabi has introduced a groundbreaking AI language model, Falcon Mamba 7B, which is the first of its kind to use State Space Language Model (SSLM) architecture rather than the traditional transformer-based approach. This new model outperforms other leading models like Meta’s Llama 3.1 8B and Mistral's 7B on key benchmarks, as verified by Hugging Face. SSLMs are known for their efficiency in processing complex, evolving information, such as entire books, without needing additional memory, making them particularly effective for tasks involving long sequences of data. The Falcon Mamba 7B, which builds on the success of TII’s previous Falcon models, represents a significant advancement in AI research and development, reinforcing Abu Dhabi’s position as a global hub for AI innovation. The model is available as open-source under a permissive license, further contributing to its impact on the AI community.

🧑🏽‍💻 AI Jobs

🛠️ AI tools updates

Tabnine, a leading AI-powered coding assistant, has expanded its capabilities with a range of new features and model options aimed at enhancing the software development process. The latest update, Tabnine Protected 2, now supports over 600 programming languages and frameworks, offering context-aware suggestions and personalized model interactions for developers. Tabnine's tools cover the full software development lifecycle, with features like inline code completions, chat-based assistance, and autonomous test generation. Notably, Tabnine's proprietary models are designed to protect user privacy and mitigate legal risks related to IP violations, making it particularly appealing to industries with stringent data security requirements. Although it currently lacks support for command-line interfaces, Tabnine's integration with popular IDEs and its ability to handle complex coding tasks position it as a strong competitor to other coding assistants like GitHub Copilot and Amazon Q Developer.

💵 Venture Capital updates

Europe's AI startup ecosystem is experiencing rapid growth, with companies expanding their teams significantly over the past year, driven by substantial funding rounds. Notable startups include Mistral AI, which has grown its team by 378% and raised nearly €1 billion to develop generative AI models; AutogenAI, which has seen a 366% increase in headcount as it automates bid writing for enterprises; and Kittl, whose team has expanded by 242% while offering AI-powered design tools. Other fast-growing companies include Qdrant, which provides an open-source search tool for AI applications, and Tacto, which helps SMEs manage their supply chains using AI. These companies are not only hiring rapidly but are also securing large investments, positioning Europe as a key player in the global AI landscape.

🫡 Meme of the day

⭐️ Generative AI image of the day