The Alignment Problem Explained: Crash Course Futures of AI #4

Could a robot dedicated to a good cause end up destroying the world? Well, maybe. In this episode, we explore how powerful AI could end up causing us harm, regardless of what it’s programmed for. Between misuse by humans, alignment problems, and instrumental goals, even AI built with good intentions could end up breaking bad…unless we do something about it. About This Series: AI is changing FAST so rather than doing a full Crash Course series of 12+ episodes, we’ve prepared a mini-series of just the basics. Crash Course will never tell you what to think and we’re not the type of organization that responds to breaking news in real time. Instead, we’re here to offer a zoomed-out foundation upon which to base your own opinions as you continue to learn from other outlets about the world that’s changing around us. Crash Course: Futures of AI will cover: -What even is AI? What’s the history of this thing and how quickly has it evolved to what exists today? -How could AI transform society? Will AI cause the next Industrial Revolution, and what might that mean for workers and the environment? -How powerful could AI become? How do we measure the progression of AI? What are the consequences we’re already seeing, and what might be the future consequences of unchecked AI development? -How might powerful AI cause harm? We’ll touch on copyright infringement, misinformation, surveillance, authoritarianism, and (unfortunately,) more. -How could AI be governed? What are the potential approaches for controlling AI both nationally and internationally? P.S. Wondering if we used AI to create this series? Nope! Every Complexly video is lovingly, painstakingly human-made. Intro 00:00 Chapter 1: How powerful AI systems can be misused 1:46 Chapter 2: Misalignment in advanced AI systems 3:36 Chapter 3: How instrumental goals could lead to rogue AI 6:43 Chapter 4: The case for the precautionary principle 9:30 Conclusion 11:49 Sources: https://docs.google.com/document/d/16br0e73KFVD5qu-VEBV40yQTpBiX0l65EgWQ6Q7Fz-k/edit?usp=sharing