Google TechTalks
Google TechTalks · 292 个带字幕的视频Google TechTalks 在 WinkLingo 收录了 292 个视频(这个频道有 35万 个订阅),每个视频都能逐词检索。
查看 YouTube 原频道英文简介
Google Tech Talks is a grass-roots program at Google for sharing information of interest to the technical community. At its best, it's part of an ongoing discussion about our world featuring top experts in diverse fields. Presentations range from the broadest of perspective overviews to the most technical of deep dives, on topics well-established to wildly speculative. Disclaimer: Google TechTalks are designed to disseminate a wide spectrum of views on topics including Current Affairs, Science, Medicine, Engineering, Business, Humanities, Law, Entertainment, and the Arts. The views or…
收录的视频
58:08本地节点差分隐私
Local Node Differential Privacy
57:54通过 b-Min-Sep 子采样实现相关噪声机制的隐私增强
Privacy Amplification for Correlated-Noise Mechanisms via b-Min-Sep Subsampling
40:17差分隐私的表格-图像多模态数据生成
Differentially Private Table-Image Multimodal Data Generation
56:22从原子事实到结构化知识:重新思考大语言模型中的“取消学习”与“越狱”
Atomic Facts to Structured Knowledge: Rethinking Unlearning & Jailbreaking in Large Language Models
18:32理解最差情况差分隐私保证与真实威胁模型之间的权衡
Understanding Trade-offs Between Worst-case Differential Privacy Guarantees & Real Threat Models
53:30机器文本检测器本质上是成员推理攻击
Machine Text Detectors are Membership Inference Attacks
57:57代码健康守护者 (Code Health Guardian)
Code Health Guardian
47:17回顾与前瞻:LLM 隐私与投毒中出现的新(旧)威胁
Going Back and Beyond: Emerging (Old) Threats in LLM Privacy and Poisoning
33:40POPri:使用偏好优化合成数据的私有联邦学习
POPri: Private Federated Learning using Preference-Optimized Synthetic Data
33:23利用单实例隐私实现机器卸载 (Machine Unlearning)
Leveraging Per-Instance Privacy for Machine Unlearning
50:30大语言模型的私有化适配
Private Adaptations of Large Language Models
40:16医疗 AI 带来的差异化隐私风险——患者级隐私风险调查
Disparate Privacy Risks from Medical AI - An Investigation into Patient-level Privacy Risk
54:48用于快速私有统计的稳定估计量
Stable Estimators for Fast Private Statistics
57:39语言模型的最坏情况成员推理 (Membership Inference)
Worst-Case Membership Inference of Language Models
38:13语言模型中从注入到蒸馏的级联对抗偏差
Cascading Adversarial Bias from Injection to Distillation in Language Models
45:47单次运行审计(One Run Auditing)的局限性与可能性
The Limits and Possibilities of One Run Auditing
33:10大语言模型 (LLM) 的持久预训练中毒问题
Persistent Pre-Training Poisoning of LLMs
53:45大语言模型的隐私审计
Privacy Auditing of Large Language Models
32:31语言模型中从注入到蒸馏的级联对抗偏差
Cascading Adversarial Bias from Injection to Distillation in Language Models
45:45无需训练的差分隐私合成数据
Differentially Private Synthetic Data without Training
49:31记忆的威胁模型:隐私、版权及其之间的一切
Threat Models for Memorization: Privacy, Copyright, and Everything In-Between
42:03语言模型训练中添加或删除个人信息带来的隐私连锁反应
Privacy Ripple Effects from Adding or Removing Personal Information in Language Model Training
51:48通过分桶实现私有连续计数的流式传输
Streaming Private Continual Counting via Binning
51:27追寻常数及其在差分隐私中的影响
Chasing the Constants and its Implications in Differential Privacy