• AI KATANA
  • Posts
  • Japan to help Southeast Asia develop AI in local languages

Japan to help Southeast Asia develop AI in local languages

Also: Runway’s Gen-3 Alpha AI video model now available – but there’s a catch

Good morning,

In today’s newsletter, we explore Japan’s initiative to develop AI in local languages across Southeast Asia, mitigating reliance on foreign tech and preserving cultural diversity. We also delve into RunwayML’s new Gen-3 Alpha AI video model, Google’s rising emissions due to AI, and the surge in AI startup investments despite a tech downturn. Additionally, Tencent’s strides in AI training efficiency without advanced Nvidia chips, Cloudflare’s anti-web-scraping tool, and Perplexity’s Pro Search AI upgrade highlight ongoing advancements in AI tools. Finally, we look at Revolut’s Nik Storonsky launching a $200 million AI-driven VC firm, aiming to transform venture capital with a data-centric approach.

Sliced:

  • 🇯🇵 Japan to help Southeast Asia develop AI in local languages

  • 🆕 Runway’s Gen-3 Alpha AI video model now available – but there’s a catch

  • 🚨 Google’s emissions shot up 48% over five years due to AI

  • 💵 Investors Pour $27.1 Billion Into A.I. Start-Ups, Defying a Downturn

  • 📈 Tencent boosts AI training efficiency without Nvidia’s most advanced chips

Japan is set to assist Southeast Asian nations in developing AI technology tailored to local languages through a new public-private initiative announced by Prime Minister Fumio Kishida. This effort, which will be detailed at the Asia Business Summit, aims to mitigate economic security risks posed by dependence on foreign technologies and to preserve cultural diversity. By fostering partnerships between Japanese AI firms and businesses across the region, the initiative will focus on training LLMs in languages spoken in countries like Singapore, Malaysia, and Vietnam. The collaboration will involve compiling text and voice data and testing these models, supported by Japan’s provision of computational resources. Notable projects include Elyza’s Thai LLM and Singapore’s efforts on Indonesian, Malay, and Thai models. Japan will also offer subsidies and support digital startups through programs like the Generative AI Accelerator Challenge and a skills-building plan targeting 100,000 individuals over five years.

RunwayML has launched its Gen-3 Alpha model, allowing users to create hyper-realistic AI-generated videos from text, image, or video prompts. Unlike its predecessors, Gen-3 Alpha requires a paid subscription starting at $12 per month per editor, billed annually. The model offers significant improvements over Gen-1 and Gen-2, including enhanced speed, fidelity, consistency, and motion capabilities, achieved through collaboration with a team of research scientists, engineers, and artists. Initially supporting text-to-video mode, the model will expand to image-to-video and video-to-video modes, leveraging advanced control features like Motion Brush and Director Mode. Each video generation is limited to 10 seconds, which, while shorter than OpenAI’s upcoming Sora model, still represents a substantial advancement in AI video generation technology. This rollout marks the beginning of a series of new models from RunwayML, aimed at simulating a wide range of real-world interactions and scenarios.

Google’s greenhouse gas emissions surged by 48% over five years, reaching 14.3 million metric tons of CO2 equivalent in 2023, primarily due to the increased integration of AI across its products. This rise, driven by higher energy consumption at data centers and emissions from its supply chain, complicates Google’s goal of achieving carbon neutrality by 2030. The resource-intensive nature of generative AI, which demands significant computational power, is a key factor in this trend. Similar challenges are faced by other tech giants like Microsoft, which also reported increased emissions due to AI investments. The unexpected energy demands of AI development have blindsided tech companies, underscoring the difficulty of aligning rapid technological advancements with stringent environmental targets.

Despite a broader downturn in the tech industry, AI startups in the United States have been thriving, with investors pouring $27.1 billion into these companies from April to June 2024. This investment accounted for nearly half of all U.S. startup funding during this period, reflecting a significant shift in venture capital focus. Major funding rounds included $1.1 billion for CoreWeave, a cloud computing service provider for AI, and $6 billion for Elon Musk’s xAI. The resurgence of investment interest follows the release of OpenAI’s ChatGPT in late 2022, which sparked renewed enthusiasm in generative AI technologies capable of creating text, images, and videos. Despite the high costs associated with AI development, including substantial expenditures on computing resources, the potential of AI has generated considerable hype, with investors betting on its future impact to surpass that of smartphones and social media. However, the intense competition from tech giants like Microsoft and Amazon may affect future funding dynamics.

Tencent has significantly enhanced the training efficiency of its AI models by optimizing its existing high-performance computing (HPC) network, known as Xingmai, instead of relying on Nvidia’s most advanced chips. This strategic upgrade, which boosts network communication efficiency by 60% and large language model (LLM) training efficiency by 20%, comes amid stringent US export restrictions on advanced semiconductors. Tencent’s improvements allow better utilization of idling GPU capacity, supporting over 100,000 GPUs in a single computing cluster. This advancement is part of a broader push by Chinese tech companies to achieve technological self-reliance and reduce costs in a competitive AI market. Tencent’s efforts also include promoting its in-house LLMs and assisting other companies in building their models, contributing to a price war among Chinese AI firms. This push for efficiency reflects a growing trend where AI companies, constrained by resource availability, are finding innovative solutions to enhance performance and reduce costs.

🛠️ AI tools updates

Cloudflare has introduced a new feature allowing web hosting customers to block AI bots from scraping website content with a single click. This measure addresses widespread customer dissatisfaction with AI bots harvesting data without permission to train machine learning models. While the traditional method of using a robots.txt file to prevent web crawlers is still available, its effectiveness is limited as bots can easily ignore it. Cloudflare’s new tool leverages machine learning to detect and block these bots more robustly, even when they use spoofed user agents to mimic real browsers. This feature, accessible even to free-tier customers, aims to safeguard content creators’ rights and maintain control over how their content is used.

Perplexity has significantly enhanced its Pro Search AI tool, improving its ability to handle complex queries, perform advanced math, and conduct thorough research. This upgrade allows the AI to break down multifaceted questions into detailed, step-by-step responses, providing comprehensive answers. Additionally, the tool now features advanced programming capabilities, enabling it to analyze data, debug code, and generate detailed reports through a new feature called Pages. Despite these advancements, Perplexity faces scrutiny over ethical concerns, with accusations of bypassing robots.txt rules to scrape data without permission. The upgraded Pro Search offers five free searches per day, with a premium option providing 600 searches daily for $20 per month.

💵 Venture Capital updates

Nik Storonsky, the billionaire behind Revolut, has launched a $200 million venture capital firm that is heavily integrated with AI to create a systematic investment process. The firm aims to revolutionize venture capital by leveraging AI for decision-making, thereby enhancing the precision and efficiency of investments. This new venture seeks to address traditional VC inefficiencies by using advanced data analysis and machine learning algorithms to identify and support high-potential startups. Storonsky’s approach emphasizes a data-driven methodology, contrasting with conventional intuition-based investment strategies, and reflects his commitment to pushing technological boundaries in the financial sector.

🫡 Meme of the day

⭐️ Generative AI image of the day