AI News Roundup – March 07, 2025
Chinese scientists build world’s first AI chip made of carbon and it’s super fast
Chinese scientists have created the world’s first carbon-based AI microchip, potentially overcoming the limitations of traditional silicon-based chips and paving the way for next-generation computing
Anthropic’s Recommendations to OSTP for the U.S. AI Action Plan
Anthropic has submitted recommendations for a U.S. AI Action Plan, emphasizing the need to strengthen national security and promote prosperity through a robust AI development ecosystem
Hugging Face co-founder Thomas Wolf just challenged Anthropic CEO’s vision for AI’s future — and the $130 billion industry is taking notice
AI leaders debate whether AI can innovate, with Thomas Wolf arguing AI prioritizes conformity over creativity
Experts don’t think AI is ready to be a ‘co-scientist’
Scientists are skeptical of claims that AI can dramatically accelerate scientific discovery, arguing current tools lack detail and empirical support
Pioneers of Reinforcement Learning Win the Turing Award
AI pioneers Andrew Barto and Rich Sutton have won the Turing Award for their work in reinforcement learning, a key technique behind modern AI advancements
Trump wants to kill $52.7 billion semiconductor chips subsidy law
Trump wants to repeal the CHIPS Act, redirecting funds from semiconductor subsidies to debt reduction, arguing it’s unnecessary spending. This creates industry uncertainty
It’s Not Just Us: AI Models Struggle With Overthinking
Researchers have discovered that advanced AI models can get stuck in “analysis paralysis”, hindering their ability to solve problems efficiently
Southeast Asia must take the long view to survive US-China AI race
An opinion piece highlights the risks of the US prioritizing AI innovation over safety, particularly in Southeast Asia, and calls for a more balanced approach that considers the interests of all nations
Chain of Draft approach allows AI models to carry out tasks using far fewer resources
Zoom researchers have developed a new AI training method called Chain of Draft (CoD) that improves efficiency and reduces computational costs compared to existing methods
Data Science Agent in Colab: The future of data analysis with Gemini
Google has released the Data Science Agent in Colab, powered by Gemini, which automatically generates executable Colab notebooks from natural language descriptions, simplifying data analysis workflows

