Technology

Core viewpoint: The focus of AI is shifting towards solving practical problems and exploring diverse entry points.

AI's Value Lies in Solving Practical Problems

The true value of AI lies in its ability to solve tangible, real-world problems.

The Battle for AI Entry Points

The competition for AI entry points is intensifying.

Learning from History Beyond the AGI Narrative

It's crucial to look to historical experiences rather than just focusing on the narrative of Artificial General Intelligence (AGI).

Technology

Core viewpoint: OpenAI's stance on AI regulation and international collaboration raises concerns about fairness and the potential for hindering global technological progress.

OpenAI's Controversial AI Policy Proposal

OpenAI is facing criticism for advocating that the U.S. government restrict the development of AI technology in China, while simultaneously lobbying for relaxed AI regulations in the U.S. This perceived double standard, specifically targeting Chinese AI institutions like DeepSeek, raises concerns about hindering global AI ecosystem diversity and slowing technological advancement.

Technology

Core viewpoint: A new diffusion model is challenging the dominance of traditional transformer models, promising faster and more efficient AI processing.

InceptionLabs Launches Mercury Diffusion Model

InceptionLabs has launched Mercury, the world's first commercial-grade diffusion model, which is designed to disrupt the traditional Transformer route. MercuryCoder surpasses models like GPT-4oMini and Claude3.5Haiku in both speed and quality, achieving a processing speed of over 1000 tokens per second on NVIDIA H100 chips. Unlike the chain-of-thought, sequential generation of Transformers, Diffusion directly generates an answer from a blurred state, offering faster computation. The emergence of Mercury brings new possibilities to AI development and may change the AI landscape.

Technology

Core viewpoint: AI data privacy and security concerns are escalating, with potential geopolitical implications.

Data Privacy Concerns in AI Chatbots

Research indicates that Google Gemini is the most data-intensive AI chatbot, collecting 22 out of 35 user data types, including sensitive information. DeepSeek ranks fifth among popular applications. Gemini, Copilot, and Perplexity collect precise location data, and about 30% of chatbots share sensitive user data with third parties.

OpenAI's Stance on DeepSeek

OpenAI has labeled Chinese AI lab DeepSeek as state-subsidized and controlled, suggesting the U.S. government consider banning its models. OpenAI claims DeepSeek's models are insecure due to Chinese law requiring compliance with user data demands, posing privacy and security risks, including IP theft. There is no clear link between the Chinese government and DeepSeek, though the PRC has taken increased interest in DeepSeek in recent months.