China is making big strides in AI, showcased at Shanghai's World AI Conference with advanced humanoid robots from companies like Unitree and Deep Robotics that can box and do backflips. Unitree plans to launch its R1 humanoid robot for under $6,000. Baidu is also developing 'digital humans' with decision-making capabilities. Premier Li Qiang announced a new group to focus on AI regulations, reflecting China's ambition to lead in the $7 trillion humanoid robot market predicted by Citigroup. Meanwhile, Z.ai, a Chinese startup, introduced its GLM-4.5 AI model, claiming it's cheaper to use than DeepSeek and requires only eight Nvidia H20 chips. In the US, Meta is facing a lawsuit from Strike 3 Holdings, an adult content company, alleging that Meta used pirated videos to train its LLaMA AI models. Strike 3 Holdings accuses Meta of copyright infringement and seeks damages. Mark Cuban suggests a ban on advertising within AI models to prevent biased recommendations, advocating for clear labeling if ads are present. Elsewhere, Google is partnering with the Indian Institute of Mass Communication (IIMC) to launch the Google News Initiative AI Skills Academy, a 10-week program to train Indian journalists and students on AI tools like NotebookLM and Gemini. This initiative aims to enhance newsroom efficiency and storytelling. Dropzone AI, a Seattle startup, secured $37 million for its AI security software, which uses AI agents to act as security analysts. As AI's power demands grow, tech companies are seeking creative energy solutions, while best practices like checklists and workflows are becoming essential for ensuring accuracy in complex e-discovery processes. Finally, graduate students are encouraged to engage with AI to shape its societal impact, using tools like ChatGPT and Beautiful.ai to enhance research and presentations.
Key Takeaways
- China is rapidly advancing in AI, demonstrated by humanoid robots at the World AI Conference in Shanghai.
- Unitree plans to launch its R1 humanoid robot for under $6,000.
- Z.ai's GLM-4.5 AI model claims lower usage costs than DeepSeek and uses only eight Nvidia H20 chips.
- Meta is being sued by Strike 3 Holdings for allegedly using pirated adult content to train its LLaMA AI models.
- Mark Cuban proposes banning ads in AI models or requiring clear labeling to avoid biased recommendations.
- Google and IIMC are launching the Google News Initiative AI Skills Academy to train Indian journalists in AI.
- Dropzone AI raised $37 million for its AI security software that uses AI agents as security analysts.
- Tech companies are seeking creative solutions to meet the growing power demands of AI.
- Checklists and workflows are crucial for ensuring accuracy in complex e-discovery processes.
- Graduate students should engage with AI to shape its impact and develop essential AI skills.
China's AI ambition shown through humanoid robots at Shanghai conference
At Shanghai's World AI Conference, dozens of humanoid robots showed China's growing AI skills. Premier Li Qiang said China will create a new group to work together on AI rules. Yang Yifan from Transwarp said there is high demand for AI in China. Unitree announced it will launch a full-size humanoid robot, the R1, for under $6,000. Baidu also announced new tech for its 'digital humans' that can think and make decisions.
Robots boxing and backflipping at China's AI summit in Shanghai
China's biggest AI conference in Shanghai featured robots that could box, serve drinks, and even do backflips. Companies like Unitree and Deep Robotics showed off their advanced humanoid robots. Unitree teased its Mini robot hanging up a shirt. Chinese companies are quickly improving in AI, even though integrating robots into daily life may still be far off. Citigroup Inc. thinks the humanoid robot market could be worth $7 trillion by 2050, and China wants to lead the way.
Meta sued for allegedly using pirated porn to train AI
Meta is being sued by Strike 3 Holdings, an adult content company, for allegedly using pirated videos to train its AI models. Strike 3 Holdings claims Meta has been downloading and sharing its copyrighted content since 2018. They say Meta did this to quickly get large amounts of data, potentially distributing the videos to minors. Strike 3 Holdings wants Meta to pay damages, stop pirating videos, and delete any stolen videos from its AI training data. Meta denies the claims.
Meta faces copyright lawsuit over adult content used in AI training
An adult content group has filed a lawsuit against Meta. The lawsuit claims Meta used the group's content without permission. The content was allegedly used to train Meta's LLaMA AI models.
Mark Cuban wants ban on ads in AI models
Mark Cuban suggests the Trump administration should ban advertising in AI models. He believes AI algorithms should not be designed to maximize revenue. Cuban worries that AI might recommend brands without users knowing if the AI is being paid. He says if AI models have ads, they should be clearly labeled and separate from user chats. Trump's administration has proposed withholding funds from states with strict AI rules.
Google and IIMC launch AI training for Indian newsrooms
Google and the Indian Institute of Mass Communication (IIMC) are starting the Google News Initiative AI Skills Academy. This 10-week program will teach AI skills to journalists and students. Participants will learn how to use Google's AI tools like NotebookLM and Gemini. The goal is to help newsrooms work more efficiently and create better stories. The program will also train media teachers and students at IIMC's six campuses.
Checklists and workflows help ensure accuracy in e-discovery
As e-discovery becomes more complex, using checklists and workflows is important. These tools help ensure completeness and accuracy when finding electronic information. The EDRM model is a good framework, but standardized processes are needed at each step. Checklists can help identify data sources and mobile device policies. Workflows should address data processing issues like deduplication and email threading. Document review benefits from planned workflows, especially with AI and predictive coding.
Big tech seeks creative solutions for AI's growing power needs
Tech giants are investing heavily in AI, but face challenges in the real world. Shortages of chips and data-center equipment are causing prices to rise and delays. Access to energy is also a problem, as utilities struggle to meet Silicon Valley's demands. President Donald Trump said America's energy issues threaten its AI leadership. Big tech companies are looking for creative ways to deal with the increasing need for power.
Dropzone raises $37M for its AI security software
Dropzone AI, a Seattle startup, raised $37 million to grow its AI security software. The company's AI agents act like security analysts, handling tasks and escalating incidents. These agents use large language models to mimic expert security analysts. Dropzone has over 100 enterprise customers, including UiPath and Zapier. The company offers a free test drive of its product and recently launched a browser extension called Dropzone AI Assistant.
Why grad students should pay attention to AI
Graduate students should engage with AI to shape its impact on society. AI is changing every field, and students need to develop AI skills to succeed. They should use AI tools thoughtfully and critically. Tools like ChatGPT, Elicit, and Perplexity can help with research and thinking. Visual communication tools like Beautiful.ai and Gamma can help with presentations. Staying informed about AI developments is also important.
New Chinese AI model claims cheaper use than DeepSeek
Z.ai, a Chinese startup, announced its new GLM-4.5 AI model will cost less to use than DeepSeek. The model is open source and free to download. GLM-4.5 uses 'agentic' AI, breaking tasks into smaller parts for better accuracy. It only needs eight Nvidia H20 chips to operate. Z.ai will charge 11 cents per million input tokens and 28 cents per million output tokens, cheaper than DeepSeek's rates.
Sources
- Humanoid robots embodiment of China's AI ambitions
- Boxing, Backflipping Robots Rule at China’s Biggest AI Summit
- Meta pirated and seeded porn for years to train AI, lawsuit says
- Meta faces adult content AI copyright lawsuit
- Mark Cuban said the Trump administration needs to crack down on ads in AI models
- Google, IIMC launch AI training program to future-proof Indian newsrooms
- Discovery beyond the basics: using checklists and workflows to ensure defensibility
- How big tech plans to feed AI’s voracious appetite for power
- Seattle startup Dropzone raises $37M to supercharge its ‘AI SOC analyst’ security software
- Why Grad Students Can’t Afford to Ignore AI (opinion)
- China's latest AI model claims to be even cheaper to use than DeepSeek