Founder, Montreal AI Ethics Institute & AI Ethics Researcher, Canada
Title: Ethical AI in Computer Graphics and Computer Vision: How do we balance progress with negative outcomes?
Biography: Abhishek Gupta
When using AI techniques to build computer graphics applications that have interactions with users in real-world contexts, it is important to eke out biases in the underlying distribution of the dataset. Neglecting to address this can lead to unwanted ethical impacts on the users, furthermore even for applications that are present in contained settings, computer games as an example, generative content from AI techniques can lead to negative consequences in terms of the algorithm picking up biases from the training datasets and exhibiting them. With examples like DeepFakes, the ethical consequences are even more dire when advances in such research can impact the lives of people in a significant manner. In the case of computer vision, the ethical implications are more evident, especially thinking about adversarial examples used to trick self-driving systems which can potentially harm human life.
This talk will illustrate examples where biased datasets and other examples can lead to unethical consequences and propose ideas on how to balance ethics with progress when deploying AI techniques in the field of computer graphics and computer vision.