Recent advances in AI and machine learning have helped create new tools and products and pushed scientific knowledge forward. They also bring along risks and complexities that we don’t yet fully understand—and these range from the hyperlocal (for example, companies perpetuating bias in their AI-powered hiring processes) to the existential (a general Artificial Intelligence wiping out life as we know it 😱 ).
It can be hard to keep up with the state of the field, let alone understand the deep implications of new research. But we’re here to help: the second season of the TDS Podcast, hosted by Jeremie Harris, focuses on these emerging questions around AI, safety, and ethics. While we encourage you to explore the entire season (now with more than two-dozen episodes, and counting!), below we’ve assembled an abridged curriculum of eight episodes that cover some of the most essential elements in this conversation.
You can approach this selection in order, or mix-and-match as you see fit. Whether you listen to it on the beach, while cooking dinner, or at your desk, we hope these discussions inspire you to keep learning and thinking about this timely, urgent topic.
Brian Christian on The Alignment Problem
A comprehensive introduction to the big questions around building safe AIs – AIs that are aligned with our values—and how it involves finding ways to very clearly and correctly quantify what we want our AIs to do.
Helen Toner on the Strategic and Security Implications of AI
As AI technology continues to develop and become more powerful, we’ll have to worry more about safety and security. But competitive pressures risk encouraging companies and countries to focus on capabilities research rather than responsible AI development.
David Roodman on Economic History and the Road to the Singularity
It’s time we explore how, and even when, transformative AI may change contemporary economies in fundamental ways.
Edward Harris on Emerging Problems in Machine Learning and Making AI "Good"
A single data scientist can design an algorithm that within minutes of its deployment will instantly affect the lives of millions, or even billions of people. It’s not so clear that we’ve become wise enough to wield such power.
Margot Gerritsen on Whether AI Have to Be Understandable to Be Ethical
Unintentionally bad AIs can lead to various biases that make algorithms perform better for some people than for others, or more generally to systems that are optimizing for things we actually don’t want in the long run.
Rob Miles on Why We Should Care about AI Safety
Should we try to ensure that advanced AI systems are perfectly safe before deploying them, or can we fix bugs as they come up? as systems become more powerful – and as their deployment is accompanied by an inevitable and irreversible loss of human control over the world – it seems likely that more caution will be needed upfront.
Silvia Milano on the Ethical Challenges of Recommender Systems
Recommender systems are incredibly valuable pieces of technology, but they also have some serious ethical failure modes – many of which arise because companies tend to build recommenders to reflect user feedback, without thinking of the broader implications these systems have for society and human civilization.
Andy Jones on AI Safety and the Scaling Hypothesis
As GPT-3 shows, when AI systems get larger, they not only perform better at the specific tasks they were trained on, but often also on completely new, more general tasks.
For the latest episodes of the Tds Podcast, visit this page.
