Redefining Ethics and Privacy in the Age of AI
The session ‘Redefining Ethics and Privacy in the age of AI’ provides a holistic view of the need to change the current outlook on privacy and ethics with the wide use and deployment of various machine learning and deep learning technologies. This session covers the ways in which privacy can be lost in machine learning and deep learning deployments and how differential privacy can be used as a privacy preserving mechanism. It also explores the possibility of using synthetic databases with privacy preservation. Lastly, it includes the ethical challenges that are currently being faced and will be faced in the future.
Outline/Structure of the Talk
The session consists of three main sub-sessions:
This sub-session provides a deep dive into the areas where privacy can be lost in ML/ DL deployments and a brief overview of how differential privacy can be used as a privacy preserving mechanism. The objective of this sub- session is to introduce the need for a new way of thinking about privacy with the vast deployment of AI solutions, explore the different areas of privacy leakage and to introduce differential privacy as a method as privacy preservation. (About 5 minutes)
Generative Adversarial Networks for Privacy:
This sub-session talks about the possibilities of using Generative Adversarial Networks to create synthetic databases with privacy preservation. The objective of this sub-session is to introduce Generative Adversarial Networks and highlight the ways in which they can be used for generating differentially private synthetic databases. It also makes a comparison between the DP GAN and PATE GAN. (About 5 minutes)
Ethics in AI:
This sub-session deals with the current ethical challenges we face with the introduction of Artificial Intelligence and explores the future challenges that we might face. The objective of this sub-session is to talk about the ethical challenges that we currently face with the vast deployment of AI solutions, the ethical challenges that we would face in the future with the advent of Artificial General Intelligence and the moral status of machines. (About 5 minutes)
In this session we expect to educate people about the various privacy risks that Machine Learning and Deep Learning possess and how to mitigate the risks. We also want to ensure that people are aware of the ethical challenges we face and will face in the future.
The key takeaways for the attendees are:
The reasons why it is important to rethink what privacy means
The ways in which privacy can be compromised
How differential privacy prevents privacy leakage
How to use differential privacy as a privacy preserving mechanism in Machine Learning / Deep Learning
What are Generative Adversarial Networks
How to use differentially private Generative Adversarial Networks for privacy preservation
Ethics in AI
Ethics for Artificial General Intelligence
Moral Status of Machines
This is an intermediate talk. For this talk, it is expected that the audience has a basic understanding of Machine Learning/ AI. This session and the sub session focus on the privacy and security implications of the extensive AI in deploying solutions and for the same, it is expected that the audience has a background of AI/ Machine Learning. While the sub-section Ethics in AI is generic, in the sub-section of Generative Adversarial Networks for Privacy, a person who doesn’t know what neural networks are, will not be able to gain much from the session.