Karen Hao

🎹 Introducing pyannote-audio 4.0 * 1 year * 100 commits * 17 contributors 🎉 New community-1 pipeline * way better speaker counting, from 61% correct to 75% on DIHARD benchmark * smoother reconciliation with STT timestamps, with the "exclusive" variant * running locally, or hosted on @pyannoteAI cloud at cost 🆕 features for power users * faster training thanks to improved dataloaders * support for pipeline repositories that also include underlying models * CLI to download, apply, optimize, and benchmark pipelines * switch to uv Python project manager 👨‍👩‍👧‍👦 Webinar and Discord community * follow me to learn about the upcoming community-1 release webinar * join our brand new Discord community Links in comment.

249

14 Comments

Adam Gleave: "Excited to have contributed to the Singapore Consensus on Global AI Safety Research Priorities, a report highlighting the key innovations needed to build trustworthy, reliable & secure AI ⚠️ Why trust this report? The report incorporates perspectives from a wide range of experts from top industry labs (e.g. OpenAI, Anthropic, GDM, xAI, Meta), academic and non-profit institutions (e.g. MIT, Stanford, Tsinghua, METR, FAR.AI), and government AI safety institutes (from the US, UK, France, Canada, China, Japan and Korea). ✅ Areas of Mutual Interest The report highlights areas of mutual interest, where it's in developers own interest to share the information with others. For example, having consistent evaluation and risk management standards increases quality while reducing compliance costs for everyone -- a win-win. ❌ Ecosystem-wide Risks This goes beyond simply harmonizing standards, though. An airplane crash isn't just bad for that airliner and manufacturer -- it's bad for the aviation industry as a whole. This is similar for AI systems. If an enterprise AI application causes an embarrassing incident or a costly lawsuit, this reduces demand for AI systems as a whole. 🦺 Sharing Safety Innovations As a result, you'd often like to share techniques that make the systems of your competitors more trustworthy and secure. In other words, it's desirable to cooperate on safety while competing strongly on capabilities. 📃 Innovation Roadmap I was excited to see the huge range of approaches identified to improve how we build AI systems identified by the experts involved. It's great to be reminded that even if a given approach doesn't pan out, many others exist. I'm confident we can find innovative solutions to these problems if we try -- but it will require a much bigger investment from companies, governments and philanthropists to bring this to fruition. Right now, many of the approaches described in the document have only a handful of people working on them. 📢 Thanks to IMDA from the Singapore government for convening this group and producing this report, and for those on the academic expert planning and writing committee for your work!"

75

3 Comments

🤖📰 Effective YESTERDAY: China has mandated a digital watermark for all AI-generated content. https://lnkd.in/gT5tCBhA Auto translate: Article 7: When reviewing applications for listing or online release, internet application distribution platforms shall require internet application service providers to clarify whether they provide AI-generated and synthesized services. If an internet application service provider provides AI-generated and synthesized services, the [...] platform shall verify the materials related to the identification of generated and synthesized content. Article 8: Service providers shall clearly state the methods, formats, and other specifications for identifying generated and synthesized content in their user service agreements, and remind users to carefully read and understand the relevant identification management requirements. Article 9: If a user requests that a service provider provide generated and synthesized content without explicit identification, the [...] provider may, after clarifying the user's identification obligations and usage responsibilities in the user agreement, provide generated and synthesized content without explicit identification, and shall retain relevant logs, including information on the recipients, for at least six months in accordance with the law. Article 10: When users publish generated and synthesized content using online information content dissemination services, they shall proactively declare and use the identification function provided by the [...] provider to identify it. No organization or individual may maliciously delete, tamper with, forge, or conceal the identifiers for generating synthetic content [...], provide tools or services for others to engage in the aforementioned malicious activities, or infringe upon the legitimate rights and interests of others through improper identification methods. Article 11: Service providers conducting identification activities shall also comply with the requirements of relevant laws, administrative regulations, departmental rules, and mandatory national standards. Article 12: When performing algorithm filing and security assessment procedures, [...] providers shall provide relevant materials for generating synthetic content identifiers in accordance with these Measures, strengthen the sharing of identification information, and provide support and assistance for preventing and combating related illegal and criminal activities. Article 13: Violations of these Measures shall be handled by the relevant competent departments, including the Cyberspace Administration of China, the Ministry of Telecommunications, the Public Security Bureau, and the Radio and Television Administration, in accordance with their respective responsibilities and in accordance with relevant laws, administrative regulations, and departmental rules. Article 14: These Measures shall come into force on September 1, 2025.

83

3 Comments

Some Monday deals news — MongoDB is buying Voyage AI for $220 million. The deal is expected to join Voyage’s AI retrieval tools with MongoDB’s real-time operational data, helping customers build AI applications with fewer hallucinations. Read more here: 🎁 Gift Link: MongoDB Buys Voyage AI for $220 Million to Bolster AI Search https://lnkd.in/e_rhQ2if

46

▶️ OpenAI sees more partnerships in South Korea despite sovereign AI drive ▶️ OpenAI downplayed concerns about South Korea’s delayed subordinate legislation for its basic AI law, with Chief Strategy Officer Jason Kwon stressing that the company sees itself as a partner to the government in pursuing the best outcome for users. He emphasized OpenAI’s track record on privacy, security and safety while supporting South Korea’s pursuit of sovereign AI that still allows collaboration with global companies. Kwon said OpenAI chose South Korea as a location for its third Asian subsidiary to deepen partnerships, aligning with the government’s goal to become a global AI powerhouse. 🔗 Full story: https://lnkd.in/geYii_hx MLex #MLexAI OpenAI #AI #SouthKorea #ArtificialIntelligence #SovereignAI

44

3 Comments

Scoop: AI startup LangChain, which helps developers build apps using the latest AI models, is in talks to raise $100 million at a $1.1 billion valuation, according to sources familiar with the deal. IVP is leading the deal, sources said. https://lnkd.in/gfXcdsBK

28

1 Comment

SwarmAid-GPT-5 Multiagent for Crisis Response | Events at lablab.ai A multi-agent AI platform that brings together 𝐃𝐚𝐭𝐚 𝐀𝐧𝐚𝐥𝐲𝐬𝐭𝐬, 𝐌𝐞𝐝𝐢𝐜𝐬, 𝐋𝐨𝐠𝐢𝐬𝐭𝐢𝐜𝐬 𝐌𝐚𝐧𝐚𝐠𝐞𝐫𝐬, and 𝐂𝐫𝐢𝐭𝐢𝐜𝐬 as collaborative agents to coordinate smarter crisis response. 🏆 𝐇𝐚𝐜𝐤𝐚𝐭𝐡𝐨𝐧 𝐏𝐫𝐨𝐣𝐞𝐜𝐭 — 𝐇𝐚𝐜𝐤𝐚𝐭𝐡𝐨𝐧 2025 Built with ❤️ & passion during Hackathon 2025, SwarmAid is designed to assist first responders, 𝐍𝐆𝐎𝐬, and governments in saving lives with speed and precision. 🌟 𝐇𝐨𝐰 𝐒𝐰𝐚𝐫𝐦𝐀𝐢𝐝 𝐇𝐞𝐥𝐩𝐬 📡 𝐃𝐚𝐭𝐚 𝐀𝐧𝐚𝐥𝐲𝐬𝐭 𝐀𝐠𝐞𝐧𝐭 – Leverages 𝐍𝐀𝐒𝐀 feeds & satellite data to identify disaster zones and vulnerable districts in real-time. 🩺 𝐌𝐞𝐝𝐢𝐜 𝐂𝐨𝐨𝐫𝐝𝐢𝐧𝐚𝐭𝐨𝐫 – Analyzes social signals (𝐓𝐰𝐢𝐭𝐭𝐞𝐫/𝐗) to triage urgent medical needs and detect overwhelmed hospitals. 🚚 𝐋𝐨𝐠𝐢𝐬𝐭𝐢𝐜𝐬 𝐌𝐚𝐧𝐚𝐠𝐞𝐫 – Plans safe delivery routes using 𝐎𝐩𝐞𝐧𝐑𝐨𝐮𝐭𝐞𝐒𝐞𝐫𝐯𝐢𝐜𝐞/𝐎𝐒𝐑𝐌 APIs to optimize transport of aid and supplies. 🕵️ 𝐂𝐫𝐢𝐭𝐢𝐜 𝐀𝐠𝐞𝐧𝐭 – Reviews and audits all plans to identify blind spots and improve resilience of the response strategy. 🎯 𝐎𝐮𝐫 𝐌𝐢𝐬𝐬𝐢𝐨𝐧 Disasters strike without warning. SwarmAid harnesses AI-powered agents, real-world data, and geospatial insights to deliver: ✅ Faster response strategies ✅ Optimized resource allocation ✅ Smarter decision-making under pressure 🛠️ 𝐓𝐞𝐜𝐡 𝐒𝐭𝐚𝐜𝐤 ⚡ FastAPI – Backend framework ⚙️ LangChain – Agent orchestration 🧠 GPT-5 (AIML API) – Core intelligence 🌍 Leaflet.js – Interactive disaster maps 🐦 Twitter/X API – Real-time social signals 🛰️ NASA EONET – Satellite disaster data 🚚 OpenRouteService / OSRM – Route optimization ⚛️ Next.js + TailwindCSS – Frontend UI 𝐓𝐞𝐚𝐦 𝐒𝐰𝐚𝐫𝐦𝐀𝐢𝐝 – 𝐇𝐚𝐜𝐤𝐚𝐭𝐡𝐨𝐧 2025 Muhammad Hashir Awaiz  – Team Lead & AI Architect Mirza Yasir Abdullah Baig – Backend Developer Fouzia Akbar – Data Scientist  Salaar Tariq – Frontend Developer Emaan Fatima – UX Designer & Content Creator 𝐕𝐢𝐝𝐞𝐨 𝐃𝐞𝐦𝐨: https://lnkd.in/dXGTcg7R 𝐆𝐢𝐭𝐇𝐮𝐛 𝐑𝐞𝐩𝐨: https://lnkd.in/d3d3MRYU 𝐋𝐢𝐯𝐞 𝐃𝐞𝐦𝐨: https://lnkd.in/dR4sFPMD #AIForGood #DisasterResponse #MultiAgentAI #GPT5 #TeamSwarmAid

36

3 Comments

#IPOALERT (China) Following Geek+'s relatively successful IPO on HKEX this summer, there's a few more #AGV & #AMR companies queueing up to go public, including SEER Robotics and Standard Robots Revenue for the two companies (SEER + Standard) is relatively modest at <$50m, but should be contextualised from a China market perspective. Interestingly, these companies (including Geek+) rely on a single domestic market intelligence source for their IPOs... good work for them, but will the AGV & AMR market really be worth >$11bn by 2029? Especially when you consider the two companies now in the IPO queue are top 5 businesses?? STIQ Ltd's view is a bit less so perhaps not as exciting for an IPO 😃 (PS market sizing always depends on clear definitions!) What's your view? Great timing? Pre tariff impact? Market growth? Demand? More about these companies in our upcoming annual 2025 AGV & AMR Robotics report publishing in the end of November (#NOTE you get a free copy of this report courtesy of Kollmorgen – Autonomous Mobile Solutions, 3Laws and MAG (Mobile Automation Group) Industry Group of MHI) - link in comments (or ask if you cannot see it)

21

8 Comments