32 Matching Annotations
  1. Jun 2026
    1. Security and utility always have a trade-off

      大多数人认为AI安全可以通过技术手段完美解决,但作者认为安全与实用性之间存在根本性权衡。这个观点挑战了技术乐观主义,指出公司在追求AI能力的同时必然会牺牲某些安全措施,暗示AI安全问题的解决不仅仅是技术问题,更是商业决策问题。

  2. Apr 2026
    1. V3.3 achieves 70.4% in Mode A (zero-LLM), with +23.8pp on multi-hop and +12.7pp on adversarial. V3.2 achieved 74.8% Mode A and 87.7% Mode C; the 4.4pp gap reflects a deliberate architectural trade-off.

      在零LLM模式下仅比有LLM支持的模式低17.3%,这一结果令人震惊。这表明生物启发的记忆架构可能比我们想象的更强大,能够在没有大型语言模型支持的情况下保持大部分性能,挑战了'强大AI必须依赖大型模型'的主流观点。

  3. Feb 2025
  4. Aug 2023
  5. Jul 2023
    1. Based on experienceand knowledge, they try to connect fevers, headaches, runny noses, and stomachpains to various diseases. In other words, the Columbia and Microsoftresearchers wrote a groundbreaking study by utilizing the natural, obviousmethodology that everybody uses to make health diagnoses.
  6. Mar 2021
  7. Jan 2021
  8. Oct 2020
  9. Sep 2020
  10. Aug 2020
  11. Jul 2020
  12. Jun 2020