miniblog.

← Back to all posts
Dec 28, 2020 at 03:18
Training a public machine learning model on private data can leak the original training data:
Privacy Considerations in Large Language Models
Privacy Considerations in Large Language Models
Posted by Nicholas Carlini, Research Scientist, Google Research Machine learning-based language models trained to predict the next word in a senten...

Related Posts

Jan 2, 2020 at 19:10
Increasing use of YouTube for training doctors in surgery procedures! https://www.cnbc.com/2019/11/24/doctors-are-watching-surgical-procedures-on-youtube.html
Sep 17, 2019 at 19:40
Fun research demonstration of multi agent completion for machine learning training. The content is interesting, but the cute visualisations really add to the effect:
Just a moment...
May 13, 2019 at 19:08
Really interesting paper exploring adversarial inputs to ML models: https://arxiv.org/abs/1905.02175 They conclude: * It's a property of the input data, not the training * You can even train a model on non-robust features and obtain a model that works well on the original input data!
Adversarial Examples Are Not Bugs, They Are Features
Adversarial Examples Are Not Bugs, They Are Features
Adversarial examples have attracted significant attention in machine learning, but the reasons for their existence and pervasiveness remain unclear. We demonstrate that adversarial examples can be directly attributed to the presence of non-robust features: features derived from patterns in the data distribution that are highly predictive, yet brittle and incomprehensible to humans. After capturing these features within a theoretical framework, we establish their widespread existence in standard datasets. Finally, we present a simple setting where we can rigorously tie the phenomena we observe in practice to a misalignment between the (human-specified) notion of robustness and the inherent geometry of the data.