顶尖研究者、头部机构与关键主题的访谈梳理、观点解读和信源核查。 https://polysight.bigmodel-eval.xyz/voices/ zh-CN Dario 思想编年、对齐科学与 2026 治理检验；事实截止日 2026-07-25（原型口径）。 https://polysight.bigmodel-eval.xyz/voices/anthropic/ https://polysight.bigmodel-eval.xyz/voices/anthropic/ Sun, 18 Oct 2026 00:00:00 GMT 从编年史、人物与路线差异到基础设施、Agent 与 AI for Science；事实截止日 2026-07-25（原型口径）。内网直播综述材料已剔除。 https://polysight.bigmodel-eval.xyz/voices/gemini/ https://polysight.bigmodel-eval.xyz/voices/gemini/ Sun, 18 Oct 2026 00:00:00 GMT 梳理杰弗里·辛顿的技术贡献、认知观、安全立场与对数字智能的长期判断。 https://polysight.bigmodel-eval.xyz/voices/geoffrey-hinton/ https://polysight.bigmodel-eval.xyz/voices/geoffrey-hinton/ Sun, 18 Oct 2026 00:00:00 GMT 从 GPT 谱系、Ilya–Sam 分歧到安全与产品化；事实截止日 2026-07-25（原型口径）。 https://polysight.bigmodel-eval.xyz/voices/openai/ https://polysight.bigmodel-eval.xyz/voices/openai/ Sun, 18 Oct 2026 00:00:00 GMT 从 Kaplan 与 Chinchilla 两套定量框架出发，梳理幂律拟合、工程杠杆与未来数据/经验约束。 https://polysight.bigmodel-eval.xyz/voices/scaling-law/ https://polysight.bigmodel-eval.xyz/voices/scaling-law/ Sun, 18 Oct 2026 00:00:00 GMT 从卷积网络、能量模型到 JEPA，梳理 Yann LeCun 对机器智能、世界模型、开源与 AI 风险的公开论证。 https://polysight.bigmodel-eval.xyz/voices/yann-lecun/ https://polysight.bigmodel-eval.xyz/voices/yann-lecun/ Sun, 18 Oct 2026 00:00:00 GMT 梳理 Yoshua Bengio 从深度学习基础研究到 AI 安全、Scientist AI 与全球治理的技术主线及其边界。 https://polysight.bigmodel-eval.xyz/voices/yoshua-bengio/ https://polysight.bigmodel-eval.xyz/voices/yoshua-bengio/ Sun, 18 Oct 2026 00:00:00 GMT 精读唐杰对长程任务、自治智能体、自我训练与 LLM OS 的递进判断，并以公开研究和行业实践检验其路线。 https://polysight.bigmodel-eval.xyz/voices/tang-jie-agi-first-principles/ https://polysight.bigmodel-eval.xyz/voices/tang-jie-agi-first-principles/ Sun, 18 Oct 2026 00:00:00 GMT 梳理马斯克关于 AI 风险、xAI/Grok、自动驾驶与能源的公开论述；剔除内网与未授权转载图。 https://polysight.bigmodel-eval.xyz/voices/elon-musk/ https://polysight.bigmodel-eval.xyz/voices/elon-musk/ Sun, 18 Oct 2026 00:00:00 GMT 精读张小珺第 140 期对 Google DeepMind 研究员姚顺宇的公开访谈，区分其与姚顺雨（yao-shunyu）的身份，并梳理 Coding、组织与「英雄主义已过去」等判断。 https://polysight.bigmodel-eval.xyz/voices/yao-shunyu-gdm/ https://polysight.bigmodel-eval.xyz/voices/yao-shunyu-gdm/ Sun, 18 Oct 2026 00:00:00 GMT 从 ReAct 与《The Second Half》到张小珺长访谈与 CL-bench，梳理姚顺雨关于语言、Agent、评测与 Context 的公开判断。 https://polysight.bigmodel-eval.xyz/voices/yao-shunyu/ https://polysight.bigmodel-eval.xyz/voices/yao-shunyu/ Sun, 18 Oct 2026 00:00:00 GMT 从 TED 2024、World Labs 产品与 Substack 纲领文出发，梳理李飞飞对空间智能、世界模型与 AI 治理的公开论述。 https://polysight.bigmodel-eval.xyz/voices/li-fei-fei/ https://polysight.bigmodel-eval.xyz/voices/li-fei-fei/ Sun, 18 Oct 2026 00:00:00 GMT 从暗涌深访与 DeepSeek 官方论文出发，梳理梁文锋关于原创创新、开源、组织克制与 AGI 路线的公开论述。 https://polysight.bigmodel-eval.xyz/voices/liang-wenfeng/ https://polysight.bigmodel-eval.xyz/voices/liang-wenfeng/ Sun, 18 Oct 2026 00:00:00 GMT 以 CVPR 2026 VGI Workshop 公开材料为主线，梳理田永龙从对比表征到生成式数据与 VLM 的十年研究轨迹。 https://polysight.bigmodel-eval.xyz/voices/tian-yonglong/ https://polysight.bigmodel-eval.xyz/voices/tian-yonglong/ Sun, 18 Oct 2026 00:00:00 GMT 从张小珺第 138 期访谈出发，整理罗福莉关于后训练 Agent、算力 3:1:1、架构与开源战略的公开判断。 https://polysight.bigmodel-eval.xyz/voices/luo-fuli/ https://polysight.bigmodel-eval.xyz/voices/luo-fuli/ Sun, 18 Oct 2026 00:00:00 GMT 从 ResNeXt、DiT、REPA/RAE 到 AMI，梳理谢赛宁以视觉表征连接识别、生成与世界模型的研究路线。 https://polysight.bigmodel-eval.xyz/voices/xie-saining/ https://polysight.bigmodel-eval.xyz/voices/xie-saining/ Sun, 18 Oct 2026 00:00:00 GMT