Keep The Future Human (original) (raw)
This essay makes the case for why and how we should close the gates to AGI and superintelligence, and what we should build instead.
If you just want the key takeaways, go to the Executive summary. Then, Chapters 2-5 will provide some background on the types of AI systems discussed in the essay. Chapters 5-7 explain why we might expect AGI to arrive soon, and what might happen when it does. Finally, Chapters 8-9 outline a concrete proposal to prevent AGI from being built.
Total reading time: 2-3 hours
Executive summaryA high-level overview of the essay. If you're short on time, get all the main points in just 10 minutes.Chapter 1: IntroductionHow we will respond to the prospect of smarter-than-human AI is the most pressing issue of our time. This essay provides a path forward.Chapter 2: Need-to-knows about AI neural networksHow do modern AI systems work, and what might be coming in the next generation of AIs?Chapter 3: Key aspects of how modern general AI systems are madeMost of the world's most cutting-edge AI systems are made using surprisingly similar methods. Here are the basics.Chapter 4: What are AGI and superintelligence?What exactly are the world's biggest tech companies racing to build behind closed doors?Chapter 5: At the thresholdThe path from today's AI systems to fully-fledged AGI seems shockingly short and predictable.Chapter 6: The race for AGIWhat are the driving forces behind the race to build AGI, for both companies and countries?Chapter 7: What happens if we build AGI on our current path?Society isn't ready for AGI-level systems. If we build them very soon, things could get ugly.Chapter 8: How to not build AGIAGI is not inevitable – today we stand at a fork in the road. This chapter presents a proposal for how we could prevent it from being built.Chapter 9: Engineering the future: what we should do insteadAI can do incredible good in the world. To get all of the benefits without the risks, we must ensure that AI remains a human tool.Chapter 10: The choice before usTo preserve our human future, we must choose to close the Gates to AGI and superintelligence.AppendixesSupplementary information, including: Technical details around compute accounting, an example implementation of a 'gate closure', details for a strict AGI liability regime, and a tiered approach to AGI safety & security standardsAcknowledgementsA few thank-yous to people who contributed to Keep The Future Human.
Please submit feedback and corrections to taylor@futureoflife.org.