课程: Building Trustworthy AI Systems: Transparency, Explainability, and Control with ISO/IEC TR 24028

免费学习该课程！

今天就开通帐号，24,600 门业界名师课程任您挑！

AI-specific privacy threats (Clause 8.3)

AI-specific privacy threats (Clause 8.3)

课程: Building Trustworthy AI Systems: Transparency, Explainability, and Control with ISO/IEC TR 24028

开始一个月试用 购买团队方案

AI-specific privacy threats (Clause 8.3)

“

- [Instructor] To effectively train a model, scientifically significant amounts of data may be necessary, that far surpasses the amount of data previously used for other business intelligence work. Yet that doesn't mean that all use cases require large amounts of data for a proper training. A team at Stanford University used a data set of only 1000 images to create an AI system that could diagnose skin cancer. A team at MIT used a data set of only 500 images to create an AI system that could detect diabetic retinopathy in eye scans. Collected data can potentially carry sensitive material that affects privacy. In this video, we will discuss and focus on threats to AI privacy. Let's define the components of machine learning training data. Data sets represent the total amount of individual data needed to train a model. Each data set item can contain data points, representing a single observable instance, like a row of data in a table that contains information, or a record related to a…

内容