Challenges
The mission of Machine Intelligence Research Institute is to make AI smarter-than human with the aim to have a positive impact on people's life. In order to get this goal, MIRI deals with three big challenges.
Friend or Enemy
Asimov states 3 laws for the robots based on the relationship with the man according to which a robot is a Friend. How could we guarantee that the robot won't become an Enemy? MIRI must design an agent highly reliable. How can MIRI get it? Studying some mathematical basic concepts such as
- Decision Theory
- Reflection Principles
- Logical Induction
How a Robot Learns
Is it possible to teach to a robot how it should study new things? That is maybe the hardest challenge of the modern society. There are many ethical aspects to evaluate such as we must ensure that a robot accepts our values without overcoming them. MIRI deals with this theme with some papers which faces the error tolerance problem
- The Value Learning Problem
- Corrigibility
AI Smarter Than Human
The main aspect to build an AI has got many ethical implications. MIRI investigates on how we can foresee which strategies and which choices are better to take in order to avoid a robots' war. First considerations can be found in the following papers.
- The Ethics of Artificial Intelligence
- Formalizing Convergent Instrumental Goals
- Intelligence Explosion Microeconomics
