Computers have come a long way since green screens; they have evolved into complex creatures. This panel explores artificial intelligence, from advanced cooking robots to algorithms that learn your musical preferences. Do we control AI or does it control us? Are we headed towards cute trash collectors or deadly terminators?
Today AI is still limited to logical questioning and answering. By 2015 it’s estimated that question answering, semantic search, and syntatic search will all be possible. By 2020 there will be a “cradle-to-grave” mental prosthesis. These both assuming AI-learning will become more prevalent giving AI greater potential.
The goal of AI is to create robust, fully autonomous agents in the real world. Well, how can we do this?
- Build complete solutions to relevant challenge tasks Complete agents: sense, decide, and act – closed loop Challenge tasks: specific, concrete objectives
- Drives research on componenet algorithms theory: improve from experience (machine learning), interact with other agents (multiagent systems)
- A top-down, empathetical approach
What happens when we achieve this goal?
- Utopia view? (Like the Jetsons)
- Or not? (like Terminator)
Question: would you rather live 50 years ago, or 50 years in the future? It’s not really clear since the world is changing in many ways for the worse.
- Normality: What do we consider to be normal, and what will be considered normal in the future?
- Behavior: What type of biology or behavior will they have to communicate with humans?
AI metabrain: The concept of AI being placed in human brains to make “super-humans” and allow us to naturally interface with devices or with each other differently.
AAAI Presidential Panel on Long-Term AI Futures