Capabilities

Meta-reasoning

Whereas PRODIGY cannot perform meta-level reasoning due to its inability to modify its operators and its inability to add to its repertoire, MAX, with its homogeneous representation of knowledge and its declarative structure, can reason abouts its operators, modify them and add new ones as needed. The class discussed that MAX might invent and deploy monitors when it recognizes the need to detect certain important situations in the environment.

Real-time Action

A potentially significant drawback to the declarative structure of MAX's representation of knowledge is that the efficiency and speed of behaviors will decrease as new knowledge is added. This will circumvent the intention of retaining real-time responsiveness as needed in an otherwise deliberative agent.

MAX may also have difficulty with protracted tasks. Again, the declarative structure of its memory and the posited reliance on two-valued logic makes the representation of continuous variables, especially elapsed time (episodic memory), difficult or impossible. This trait is common to many architectures.

Integration of Reasoning, Learning and Reacting

Architectural support for this integration comes from the declarative knowledge and meta-reasoning capabilities.

Manipulation of Environment

As agents acting in a dynamic robotic environment, these agents are equipped with hardware to pick up and put down objects, put out fires and recharge its battery. These activities are most likely implemented in a toy world.

Adequacy

MAX has features purportedly adequate for robotic tasks.

Coherence

MAX's meta-reasoning capability is meant to extend coherence within a particular task to a context that may include multiple simultaneous goals and prioritized capabilties. If MAX sees a fire, it will put the fire out before continuing with a more mundane task such as locating a hammer.