Artificial Intelligence

AI is the study and design of intelligent agents. An intelligent agent:

• Has own thread of control, continually executing
• Is autonomous (more on autonomy in Lecture 2)
• Exhibits proactive, reactive, and social behaviour
• Can be physical or virtual, embodied or software-only

We distinguish 2 types of AI systems:

• Centralized AI systems, also known as single-agent systems.
  • Distributed computation yet centralized control!
• Collaborative AI systems, also known as multi-agent systems.
  • Decentralized computation and control

Most of the work in AI so far is about algorithms and data structures, but we shift our focus to real time systems. How can agents communicate between them and achieve consensus?

Collaborative AI

Collaborative AI is the study of Sociotechnical Systems:

Some examples of these challenges:

• Human-machine-cooperation in robot assisted surgery
• Search and rescue mission using mobile robot swarms

CAI is crucial, as for many real-world problems:

• No single agent has all information or capabilities for solving a problem
  • Data is decentralized
  • Computation is decentralized (and asynchronous)
• There is no system for global control