AI · Theory

Intelligent Agent

Theory

16/215

Intelligent Agent

An intelligent agent chooses actions that are expected to do well on its performance measure, given what it can sense. Rational ≠ omniscient. It can be rational with limited sensors. A vacuum that sucks when it sees dirt is rational for a tiny world even if it never “understands” dirt.

Don’t mix intelligent with conscious. Interview word is rational: best expected performance, not “smart like a student”.

Output — idea / do / check for Intelligent Agent. Fill those three in the viva.

Diagram

sensors  →  agent  →  actuators
                │
                ▼
            environment

Exam tip

Rational vs omniscient in one line.

Example

# Intelligent Agent
print("idea :", "Intelligent Agent")
print("do   : one tiny example on paper")
print("check: one failure case")

Output: idea / do / check for Intelligent Agent. Fill those three in the viva.

Short notes

DefRational = good expected performance.
RuleLimited sensors still OK.
TrapIntelligence ≠ consciousness.

AI · Theory

Intelligent Agent

← All stacks

Theory

16/215

Intelligent Agent

Don’t mix intelligent with conscious. Interview word is rational: best expected performance, not “smart like a student”.

Output — idea / do / check for Intelligent Agent. Fill those three in the viva.

Diagram

sensors  →  agent  →  actuators
                │
                ▼
            environment

Exam tip

Rational vs omniscient in one line.

Example

# Intelligent Agent
print("idea :", "Intelligent Agent")
print("do   : one tiny example on paper")
print("check: one failure case")

Output: idea / do / check for Intelligent Agent. Fill those three in the viva.

Short notes

DefRational = good expected performance.
RuleLimited sensors still OK.
TrapIntelligence ≠ consciousness.

Intelligent Agent

Example

Short notes

Questions

Intelligent Agent

Example

Short notes

Questions