Three generals are voting on whether to attack or retreat from their siege of a castle. One of the generals is corrupt and two of them are not. What happens when the corrupted general sends different answers to the other two generals?

A Byzantine fault is “a condition of a computer system, particularly distributed computing systems, where components may fail and there is imperfect information on whether a component has failed. …

What motivates cooperative inverse reinforcement learning? What can we gain from recontextualizing our safety efforts from the CIRL point of view? What possible role can pre-AGI systems play in amplifying normative processes?

Cooperative Inverse Reinforcement Learning with Dylan Hadfield-Menell is the eighth podcast in the AI Alignment Podcast series, hosted by Lucas Perry and was recorded at the Beneficial AGI 2018 conference in Puerto Rico. For those of you that are new, this series covers and explores the AI alignment problem across a large variety of domains, reflecting the fundamentally interdisciplinary nature of AI alignment. Broadly, Lucas will speak…

Humanity is at a turning point. For the first time in history, we have the technology to completely obliterate ourselves. But we’ve also created boundless possibilities for all life that could enable just about any brilliant future we can imagine. Humanity could erase itself with a nuclear war or a poorly designed AI, or we could colonize space and expand life throughout the universe: As a species, our future has never been more open-ended.

The potential for disaster is often more visible than the potential for triumph, so as we prepare for 2019, we want to talk about existential…

By Kirsten Gronlund

In both 2016 and 2017, genome editing made it into the annual Worldwide Threat Assessment of the US Intelligence Community. One of biotechnology’s most promising modern developments, it had now been deemed a danger to US national security — and then, after two years, it was dropped from the list again. All of which raises the question: what, exactly, is genome editing, and what can it do?

Most simply, the phrase “genome editing” represents tools and techniques that biotechnologists use to edit the genome — that is, the DNA or RNA of plants, animals, and bacteria. Though…

How can humanity survive the next century of climate change, a growing population, and emerging technological threats? Where do we stand now, and what steps can we take to cooperate and address our greatest existential risks?

In this special podcast episode, Ariel speaks with Martin Rees about his new book, On the Future: Prospects for Humanity, which discusses humanity’s existential risks and the role that technology plays in determining our collective future. Martin is a cosmologist and space scientist based in the University of Cambridge. He is director of The Institute of Astronomy and Master of Trinity College, and…

By Tucker Davey

At the core of AI safety, lies the value alignment problem: how can we teach artificial intelligence systems to act in accordance with human goals and values?

Many researchers interact with AI systems to teach them human values, using techniques like inverse reinforcement learning (IRL). In theory, with IRL, an AI system can learn what humans value and how to best assist them by observing human behavior and receiving human feedback.

But human behavior doesn’t always reflect human values, and human feedback is often biased. We say we want healthy food when we’re relaxed, but then we…

How are we to make progress on AI alignment given moral uncertainty? What are the ideal ways of resolving conflicting value systems and views of morality among persons? How ought we to go about AI alignment given that we are unsure about our normative and metaethical theories? How should preferences be aggregated and persons idealized in the context of our uncertainty?

Moral Uncertainty and the Path to AI Alignment with William MacAskill is the fifth podcast in the new AI Alignment series, hosted by Lucas Perry. For those of you that are new, this series will be covering and…

By Jolene Creighton

Our AI systems work remarkably well in closed worlds. That’s because these environments contain a set number of variables, making the worlds perfectly known and perfectly predictable. In these micro environments, machines only encounter objects that are familiar to them. As a result, they always know how they should act and respond. Unfortunately, these same systems quickly become confused when they are deployed in the real world, as many objects aren’t familiar to them. This is a bit of a problem because, when an AI system becomes confused, the results can be deadly.

Consider, for example, a…