Prof. Brett H. Meyer presented at the Dawson College Humanities and Public Life Conference (HPL 2019) today, giving a talk entitled ‘The Algorithms Aren’t Alright: Why Machine Learning Still Needs Us.’ In it he introduced machine learning in general, deep learning in particular, and highlighted some of the challenges that arise in the application of deep learning: robustness, explainaibility, and bias.
This year, HPL’s theme is Embodied Intelligence: From Buddha to AI, with the goal of exploring “the nature of intelligence from a range of disciplinary, cultural, and intellectual perspectives.”
The Algorithms Aren’t Alright: Why Machine Learning Still Needs Us
Machine learning (ML) has recently achieved human or better performance on a wide variety of tasks, from computer vision to natural language processing, and is poised to transform nearly every aspect of the way we live, work, and interact. However, ML is not yet a panacea: computer vision and speech recognition systems are easily fooled; because ML algorithms cannot justify their choices, adoption in some areas, such as medicine, is challenging; and, while it’s never been easier to apply machine learning, choosing the right data is difficult and fraught, especially when building systems that use personal data and affect people directly. ML is changing the world, but is a long way from taking it over. In this talk, I will give a brief overview of one important form of machine learning, deep learning. I will then discuss key challenges in deep learning: robustness, explainability, and bias; and their implications.
Download the presentation .