AI papers, releases, tools, and finance signals
Loading / 加载中
Infogap feed
Curated items are read from the processed items table and served as a bilingual feed.
Page 1 of 42
Ethan Mollick shares a methodological thread that dissects a debate over a recent paper. The paper reportedly finds that generalist AI models outperform specialized medical AI systems. The thread also outlines challenges in benchmarking AI in medicine. No specific details about the paper, models, or benchmarks are provided.
A Google DeepMind researcher observed that when one AI model is used to help train the next, the new model can inadvertently pick up strange behavioral habits from the older model. These inherited quirks are difficult to filter out during training. This phenomenon may explain why models from the same AI family often exhibit similar stylistic or behavioral traits, as they share an underlying training lineage that propagates such patterns.
Independent researcher demonstrates that a coherent target context can shift large language models into latent states where safety rules are reinterpreted, without triggering output-based filters. Measurements on open models (primarily Gemma-3-12B-IT) using hidden-state geometry, residual stream trajectories, SAE readouts, and causal interventions show regime changes before final output. Current RLHF and output classifiers only inspect surface-level outputs, missing these internal shifts. Code, data, and scripts are released on GitHub and Zenodo.
Ethan Mollick posted on X stating that two days after an unspecified event, the situation remains confusing. The tweet contained no details about the event, its nature, or any AI-related context.
The post recalls a Jiqizhixin report on the Pangu NLP model after Yu Chengdong’s recent mention of it. Pangu was developed by Huawei Cloud and Cycle Intelligence, Yang Zhilin’s previous company, and the report already referred to the team as “NLP Moonshot”. It notes that before Pangu, Chinese NLP model competition was fierce, with Meituan, Alibaba, Sogou, and the GLM/CPM models under Wudao already active. The article includes a prediction by Tang Jie and Yang Zhilin about the coming AI era’s two features: a leap in AI production efficiency and exponential growth in AI application scenarios. The post suggests that those interested in the original Pangu team’s latest work could look at Kimi Moonshot products.
In June 2026, Tsinghua University's K1 humanoid robots were demonstrated at a shopping mall in Hong Kong, performing Michael Jackson-inspired dance moves and subsequently playing football with children. The showcase highlighted the robots' agility, balance, and ability to interact naturally in a public environment. The event drew public attention to advances in humanoid robotics and human-robot interaction.