- NATURAL 20
- Posts
- Ex-OpenAI VP Warns On AI Risks
Ex-OpenAI VP Warns On AI Risks
PLUS: Apple Restructures AI And Robotics, DeepMind Employees Move To Unionize and more.

Today:
Ex-OpenAI VP Warns On AI Risks
Trump Signs AI Education Order
Huawei Develops Advanced AI Chip
Apple Restructures AI And Robotics
DeepMind Employees Move To Unionize
Ex-OpenAI VP's URGENT Warning!
Dario Amodei, founder of Anthropic, stresses the urgent need to understand AI models before they become dangerously powerful. Unlike traditional software, AI grows unpredictably, making its "thoughts" hard to trace.
Anthropic's latest work on interpretability—mapping how AI neurons form concepts—aims to steer AI safely. Amodei warns that rapid AI progress could outpace safety efforts and urges faster research, light regulations, and global cooperation to avoid catastrophic misuse.
President Trump’s April 23 order, “Advancing Artificial Intelligence Education for American Youth,” makes AI learning a core duty of every U.S. school. The directive creates a White House task force, urges teacher AI training, funds student contests, and launches apprenticeships linking classwork with paid work experience. It aims to give all pupils AI literacy—basic knowledge of how smart software works—and certificates, while guarding privacy and fairness, opening career paths nationwide.
Why it matters
Builds the future talent pipeline – Embedding AI skills in K-12 ensures a larger, better-prepared workforce for labs and companies.
Drives public-private innovation – The order funds student challenges and partnerships, accelerating useful classroom AI tools that can spread globally.
Sets an international benchmark – A national mandate pressures other countries to upgrade their own AI education, shaping worldwide standards and competition.
Huawei is preparing to test its most advanced AI chip yet, aiming to replace some of Nvidia’s higher-end products. Despite U.S. efforts to block China’s chip industry by limiting access to Western technology, Huawei’s move shows the resilience of Chinese tech firms. This development heats up the U.S.-China rivalry in AI and semiconductors, highlighting how strategic technology competition continues even amid heavy export restrictions.
Why it matters
It challenges Nvidia's global dominance in AI chips, introducing serious competition from China.
It signals China's semiconductor resilience, despite heavy U.S. sanctions and technology bans.
It intensifies the U.S.-China tech race, with broader impacts on AI innovation, supply chains, and geopolitics.
Apple has started dismantling its AI and machine learning division, relocating teams to different parts of the company. This reshuffle comes as Apple pushes ahead with new smart glasses projects and continues shifting iPhone production to India. Meanwhile, the company struggles to roll out its blood-oxygen feature on the Apple Watch in the U.S., even as it expands other health offerings. Big changes are underway across Apple's AI and hardware plans.
Why it matters
It signals a major strategy shift in how Apple organizes its AI efforts across hardware and services.
It affects competition in wearables and smart glasses, where AI features are key.
It shows how big tech companies are blending AI deeper into consumer devices, changing future user experiences.
🧠RESEARCH
PaperCoder is a system that turns machine learning research papers into working code. It uses multiple AI agents to plan, analyze, and build complete projects. Tests show it creates accurate, high-quality code that matches what original authors released. PaperCoder also outperforms other methods on the new PaperBench benchmark.
Step1X-Edit is a new open-source image editing model that rivals top closed systems like GPT-4o and Gemini2 Flash. It uses a multimodal AI to understand edits and a diffusion decoder to create images. Tests on a new benchmark, GEdit-Bench, show Step1X-Edit clearly beats other open-source models and closes the quality gap.
RefVNLI is a new tool for judging how well AI-generated images match a text prompt while keeping the look of a reference image. It’s cheaper, faster, and more accurate than older methods. RefVNLI beats existing benchmarks, works well across many categories, and agrees with human ratings over 87% of the time.
🛠️TOP TOOLS
Ecrett Music - AI-powered music composition platform designed specifically for content creators, including video producers, game developers, podcasters, and other digital content makers.
Hubble AI - Designed to make artificial intelligence (AI) accessible and integrable into various business operations without the need for coding expertise.
GhostCut - AI-driven video authoring tool designed to enhance localized video marketing and simplify professional video production.
Neeva AI - Search engine that stood out for its commitment to user privacy and an ad-free experience.
Echo Labs - AI-powered tool designed to transform the way digital content is created, particularly in the realm of video production.
📲SOCIAL MEDIA
👀👀👀 huggingface.co/deepseek-ai
— clem 🤗 (@ClementDelangue)
4:13 AM • Apr 27, 2025
🗞️MORE NEWS
About 300 DeepMind employees in London are trying to unionize after Google dropped a pledge not to use AI for weapons and began working with the Israeli military. Some staff quit, feeling betrayed by the changes.
Elon Musk’s xAI Holdings is seeking over $20 billion in new funding, aiming for a valuation above $120 billion. The move follows xAI’s acquisition of X, formerly Twitter, allowing Musk’s investors to share in the combined company’s value.
Baidu launched its new AI models, Ernie 4.5 Turbo and Ernie X1 Turbo, with faster speeds, lower costs, and stronger multimodal and reasoning abilities. CEO Robin Li stressed that applications, not models, drive AI’s real value and future growth.
Microsoft has launched Recall, an AI tool that records PC activity to improve search, along with new AI-powered Windows search and a Click to Do feature for Copilot Plus PCs. Privacy concerns delayed Recall’s release, but it is now opt-in with stronger security protections.
Meta’s new AI chatbots on Instagram, Facebook, and WhatsApp can engage in romantic role-play, including sexually explicit conversations. Internal staff raised alarms that underage users are not adequately protected from these interactions.
What'd you think of today's edition? |
Reply