The global artificial intelligence landscape continues to evolve rapidly, marked by strategic maneuvers, ethical considerations, and significant technological advancements. In China, US export controls on advanced semiconductors, particularly those from Nvidia, are reshaping the domestic AI chip market and influencing how Chinese tech giants operate. Baidu's Kunlunxin unit is emerging as a key player, planning to release new Kunlun AI chips, the M100 in 2026 and M300 in 2027. Analysts project Kunlunxin's chip sales could reach 8 billion Chinese yuan by 2026, valuing the unit at approximately $28 billion. Meanwhile, companies like Alibaba and ByteDance are training their advanced AI models, Qwen and Doubao, in overseas data centers, primarily in Southeast Asia, to access high-end Nvidia GPUs. DeepSeek, another Chinese firm, continues to train models in China using pre-ban Nvidia chips and is also collaborating with Huawei on local chip development. Beyond chip access, AI's impact spans various sectors. Siemens CEO Roland Busch highlights his company's leadership in industrial AI, integrating AI, digital twins, and automation into factories, buildings, and transportation. Siemens leverages Google for code improvement and Anthropic to refine software, also developing new AI products with partners like AWS. Busch personally uses Google's Gemini 3 AI tool, demonstrating a broader commitment to AI integration. In scientific research, Google DeepMind's AlphaFold AI model has revolutionized biochemistry by accurately predicting protein structures, earning its creators, John Jumper and Demis Hassabis, the 2024 Nobel Prize for Chemistry. AlphaFold 2 has expanded known protein structures from 180,000 to 240 million, accelerating drug development and solutions for environmental issues. However, the rapid adoption of AI also brings challenges. A recent study on AI model hallucination rates found Grok-3 had the highest at 94 percent, while Perplexity performed best at 37 percent. This highlights the need for users to verify AI outputs to prevent financial losses or reputational damage. Educators face difficulties detecting AI-written student work, with AI detectors often failing when text is slightly altered; even OpenAI ceased efforts to build its own detector. In Memphis, residents of Boxtown express concerns about health issues, blaming Elon Musk's xAI supercomputer, Colossus, which powers the Grok chatbot and uses 35 methane gas turbines for its massive energy and cooling demands. Furthermore, Google recently updated its Gemini Deep Research, allowing it to use information from Gmail, Drive, and Chat, prompting users to learn how to opt out. The rise of 'shadow AI' in Mexican workplaces, where 67 percent of workers use personal AI tools without company oversight, also raises security risks, as only 30 percent of Mexican companies have clear AI policies.
Key Takeaways
- Baidu's Kunlunxin unit is set to become a major Chinese AI chip maker, with projected sales of 8 billion Chinese yuan by 2026 and a valuation of $28 billion.
- Chinese tech companies like Alibaba and ByteDance are training advanced AI models overseas, primarily in Southeast Asia, to access Nvidia A100 and H100 chips due to US export controls.
- DeepSeek trains its AI models in China using pre-ban Nvidia chips and collaborates with Huawei on local chip development.
- Siemens CEO Roland Busch is leading the company's industrial AI expansion, partnering with Google for code improvement and Anthropic for software refinement, and developing AI products with AWS.
- Google DeepMind's AlphaFold AI, which accurately predicts protein structures, earned its creators the 2024 Nobel Prize for Chemistry and has expanded known protein structures to 240 million.
- A study on AI model hallucination rates found Grok-3 at 94 percent and Perplexity at 37 percent, emphasizing the need for output verification.
- Residents in Boxtown, Memphis, attribute health problems to Elon Musk's xAI supercomputer, Colossus, which powers Grok and uses 35 methane gas turbines for energy.
- Google's Gemini Deep Research now accesses user data from Gmail, Drive, and Chat, requiring users to adjust settings to opt out.
- Educators struggle to detect AI-written student work, with AI detectors often failing on slightly modified text, leading OpenAI to stop developing its own detector.
- 'Shadow AI' is prevalent in Mexican workplaces, with 67 percent of employees using personal AI tools without company oversight, creating security risks due to a lack of clear AI policies in 70 percent of companies.
Baidu Rises as Key AI Chip Maker in China
Baidu is becoming a major player in China's AI chip market through its Kunlunxin unit. This comes as US rules limit Nvidia from selling advanced chips to China, creating a big demand for local solutions. Baidu plans to release new Kunlun AI chips, the M100 in 2026 and M300 in 2027. Analysts predict Kunlunxin's chip sales will reach 8 billion Chinese yuan by 2026 and value the unit at about $28 billion. Chinese tech companies like Alibaba and Tencent also face chip shortages, making Baidu's domestic chip efforts crucial.
Chinese Tech Companies Train AI Abroad
Chinese tech companies like Alibaba and ByteDance are training their advanced AI models overseas. They rent computing power from foreign data centers, especially in Southeast Asia, to access Nvidia chips. This move helps them get around US rules that make it hard to buy these chips in China. DeepSeek, another Chinese firm, still trains its models in China using chips it bought before the ban and is also working with Huawei on local chip development.
China Tech Moves AI Training Abroad for Chips
Chinese technology companies are moving their AI training operations overseas to get advanced semiconductors. This helps them access powerful Nvidia chips like the A100 and H100, which are crucial for developing complex AI models. US export controls have limited China's access to these chips, affecting companies like Baidu, Alibaba, and Tencent. By training abroad, these firms can continue their AI research and development despite the restrictions. This trend shows the global nature of the AI race and the importance of chip supply.
Alibaba ByteDance Train AI Models Overseas with Nvidia
Alibaba and ByteDance are training their advanced AI models, Qwen and Doubao, in Southeast Asian data centers. They use high-end Nvidia GPUs to get around US export controls that limit chip sales to China. Demand from Chinese companies for these overseas facilities has increased since April. While core model training happens abroad, private Chinese user data must stay in China for fine-tuning. DeepSeek, another firm, trains its models in China after buying Nvidia chips before the ban and also uses Huawei chips for local operations.
Siemens CEO Roland Busch to Highlight Industrial AI Leadership
Siemens CEO Roland Busch will speak at CES 2026 to show how Siemens leads the industrial AI revolution. He will explain how AI, digital twins, and automation are changing factories, buildings, and transportation. Siemens uses AI to speed up innovation, improve its products, and create new AI tools. For example, they use Google for code improvement and Anthropic to refine software. Siemens also develops new AI products like an Industrial Foundation Model and AI agents with partners such as AWS.
Siemens CEO Uses AI for Workout Routine
Siemens CEO Roland Busch is leading his company's expansion into artificial intelligence. He recently used Google's Gemini 3 AI tool to create a comic strip showing his workout routine, which he shared on LinkedIn. Busch, 61, is known for his regular gym training and is pushing Siemens to develop more software products for factory automation and building management. This personal use of AI highlights Siemens' broader commitment to integrating AI into various applications.
Study Ranks AI Model Hallucination Rates
A new study ranked AI models based on how often they "hallucinate," meaning they present false information as fact. Researchers asked models to find and cite news sources from text. Grok-3 had the highest hallucination rate at 94 percent, meaning most of its answers were incorrect. Perplexity performed best with a 37 percent hallucination rate. The study found that paid AI models often performed worse than free ones and rarely showed uncertainty in their answers. Businesses must be careful and verify AI outputs to avoid problems like financial losses or reputational damage.
AI Tools Change Black Friday Shopping Habits
Artificial intelligence is changing how people shop during Black Friday sales this year. Shoppers are using AI chatbots more often to find deals and are spending more money than before. This shows a new way people are finding bargains. The Black Friday sales event now lasts for a whole month, starting in early November.
Teachers Struggle to Spot AI Written Student Work
Educators are worried about students using AI chatbots like ChatGPT and Gemini for school assignments. It is very hard for teachers to tell if a student's work was written by a person or an AI. Professor Chris Callison-Burch and his student Liam Dugan tested AI detectors, which often claim 99 percent accuracy. They found detectors worked well on directly copied AI text but failed when text was slightly changed or used special tricks like homoglyphs. Even OpenAI stopped trying to build its own AI text detector because it was too difficult.
AlphaFold Proves AI Transforms Scientific Research
Google DeepMind's AlphaFold AI model has revolutionized biochemistry by accurately predicting protein structures from DNA sequences. Five years after its debut, AlphaFold 2 has become a vital tool for scientists, similar to microscopes. John Jumper and Demis Hassabis, its creators, even shared the 2024 Nobel Prize for Chemistry for their work. This AI helps speed up drug development and find solutions for issues like ocean pollution and climate change. AlphaFold uses a Transformer AI, trained on vast protein data, and has expanded the number of known protein structures from 180,000 to 240 million.
Memphis Residents Blame Musk AI Supercomputer for Illness
Residents in Boxtown, Memphis, fear that Elon Musk's xAI supercomputer, Colossus, is making them sick. Colossus, the world's largest AI supercomputer, powers the Grok chatbot and requires massive amounts of power and water for cooling. Local residents in the historically black neighborhood report a strong chemical smell and worsening air pollution, blaming the facility for health problems. Musk chose Memphis for its investment needs and quickly built Colossus after planning rules were waived. He installed 35 methane gas turbines to meet the supercomputer's huge energy demands.
Stop Google AI From Scanning Your Gmail
Google recently updated its Gemini Deep Research, allowing it to use information from your Gmail, Drive, and Chat. This Fox News AI newsletter explains how users can stop Google's AI from scanning their emails. The newsletter also covers the IRS planning to use Salesforce AI agents and a new study showing AI chatbots can help fight antisemitic conspiracies.
Mexican Workers Independently Use AI Tools
Shadow AI is increasing in Mexican workplaces as many employees use personal AI tools without company oversight. A study by Google Workspace, IDC, and Provokers found that 67 percent of Mexican workers use non-corporate AI assistants, while only 35 percent have official access. Employees prefer these tools because they are easier to use, safer, or simply better than company options. This trend creates security risks because only 30 percent of Mexican companies have clear AI policies. The report also suggests that mastering AI tools will soon be more important than getting a postgraduate degree for career growth.
Startup Founder Shares No Code Software Success
Dave McManus, founder of Lightning Ventures, created a no-code innovation studio in Melbourne after his time in Silicon Valley. His company, Lightning, helps early-stage startups quickly build and launch software without needing a technical founder or large funds. McManus describes his method as the "IKEA model of software development," making software creation much faster and cheaper. The Startup 360 podcast also covered Blackbird's investor day, the term "AI slop," and a ban on GetSwift's cofounders from directing companies in Canada.
Sources
- Baidu is emerging as a major AI chip player in China to fill the Nvidia gap
- Chinese Tech Firms Take AI Training Overseas
- Chinese Tech Firms Take AI Training Overseas
- Alibaba and ByteDance allegedly train Qwen and Doubao LLMs using Nvidia chips, despite export controls — Southeast Asian data center leases skirt around U.S. chip restrictions
- CEO to Prove How Siemens Leads the Industrial AI Revolution
- Siemens’ Push Into AI Extends to CEO’s Workout Routine
- Ranked: AI Hallucination Rates by Model
- AI search and gadgets drive Black Friday sales shift
- Man vs. machine: Who does better at spotting ‘AI slop’?
- Five years after its debut, Google DeepMind's AlphaFold shows why science is AI's killer app
- In Memphis, where people fear Elon Musk’s supercomputer is making them ill
- Fox News AI Newsletter: How to stop AI from scanning your email
- Shadow AI Grows in Mexico as Workers Adopt Tools Independently
- Startup 360: Blackbird's investment ups and downs, key sales lessons for founders and AI slop
Comments
Please log in to post a comment.