Q: In what way that knowledge-based agents's approach to intelligence is similar to human? A: They perform reasoning on internal representation of knowledge.

Q: Two limitations of problem-solving agents have, compare to knowledge-based agents? A: Their implementations are domain-specific and the atomic representation of states.

Q: How atomic representation of states of problem-solving is a limitation? A: The only choice for representing what it knows about the state is to list all possible states, which will eventually be impractical in large environments.

Q: What to call a sentence that is taken as given without being derived from other sentence? A: axiom

Three functions that abstract the implementation details of knowledge representation language? ?

• $\text{MAKE-PERCEPT-SENTENCE}$
• $\text{MAKE-ACTION-QUERY}$
• $\text{MAKE-ACTION-SENTENCE}$

Two main function as the main means of interfacing between the agent and the knowledge base (KB)? ?

• $\text{TELL}$
• $\text{ASK}$

The details of the inference mechanisms are hidden inside $\text{TELL}$ and $\text{ASK}$.

The knowledge base (KB) could initiallly contains background knowledge.

The procedural approach encodes desired behaviors directly as program code.

In the 1970s and 1980s, advocates of two approaches, declarative and procedural engaged in heated debates.

When we need to specify only what the agent knows and in order to fix its behavior, it means we fix it at the knowledge level, not the implementation level