Positive Intelligent Technologies
Self-Aware Systems is a Palo Alto think tank working to ensure that intelligent technologies are beneficial for humanity. We are in the midst of an AI and Robotics revolution. IBM, Google, Microsoft, Facebook, Baidu, and others are spending billions of dollars per year on these technologies. Why? McKinsey predicts that these technologies will create 50 trillion dollars of value by 2025. This will drive increasing investment in coming years. Gartner predicts that one-third of current jobs will be automated by 2025.
The changes are coming in three phases: The current “AI Economy” is rapidly automating economic activities to increase productivity. The emergent “AI Military” is automating warfare and creating arms races to develop military drones, robot soldiers, swarm boats, and autonomous military vehicles. The future “AI Society” will expand legal and political institutions to regulate autonomous systems. Each phase introduces its own challenges.
Popular shows like “2001“, “Terminator”, “Her”, “Person of Interest”, “Transcendence”, “Ex Machina”, “Humans”, and “Extant” portray dystopian futures in which machines turn on us in various ways. Humanity has long been fascinated by the idea of breathing life into inanimate objects, but petrified of the consequences. The ancient myths of golems, tulpas, and genies explore this theme. In the last few years, rapid advances in AI and robotics are being met with increasing popular uneasiness. In response, researchers and technology CEOs are writing defensive articles that argue that nothing they are working on is dangerous.
But are there reasons to proceed cautiously? Of course there are! This is likely to be the largest technological transformation in human history. We must approach it carefully and deliberately and make choices with great care. Are fear and “doomsaying” useful responses? Of course not! We are building these systems. We can build them however we want as long as we have the scientific understanding and the will to do so. This is a time for resolve, not for fear. What we really should be railing against is bad and sloppy design.
A system’s goals may be chosen independently from its intelligence and can lead to either harm or good. Simplistic goals give rise to unintended drives that can be anti-social. But even “superintelligences” are limited by the laws of mathematics, physics, and cryptography. Mathematical proof can be used to constrain systems with a high confidence of safety. New cryptographic techniques can be used to create a trusted computation and communication infrastructure. Cryptocurrency technology can enable decentralized systems to cooperate without prior trust. The “Safe-AI Scaffolding Strategy” can leverage this infrastructure to create a path to human thriving.
Humanity’s greatest achievement has been to cooperate on an unprecedented scale. We did this first in groups of 150 through moral emotions, language, and cooperative social structures. About 10,000 years ago agriculture allowed us to form much larger groups. Our innovations of money, free markets, laws, and governments have enabled cooperation on an unprecedented scale. The internet, social media, and decentralized technologies are currently extending cooperation to the whole of humanity. Our challenge for the coming century is to extend it to our AI and Robotic technologies as well.
This website contains our talks and papers on these topics over the past 10 years. Our recent paper “Autonomous technology and the greater human good” was downloaded 10,000 times in its first week and is the most downloaded paper ever published by the Journal of Experimental & Theoretical Artificial Intelligence. James Barrat’s book “Our Final Invention: Artificial Intelligence and the End of the Human Era” is a popular account of some of the issues and is based in part on our work. For a one-hour summary of the issues, watch this March 2015 talk at IBM Research:
For a quick introduction, watch this 18 minute TEDx video from May 2012:
For more detail, watch this December 2012 talk at Oxford University:
this July 2011 talk at Monash University:
or this October 2007 talk at Stanford University:
For more depth, read the papers:
Send email to Steve Omohundro: