Chapter 11: AI

Question 1

Intelligent agents

Answer 1

• Autonomous goal-directed entity which observes using sensors and acts upon an environment using actuators.

Question 2

Turing test

Answer 2

• Human interrogator communicates with test subject by typewriter.
• Can the human interrogator distinguish whether the test subject is a human or a machine?

Question 3

Weak AI versus strong AI

Answer 3

• Weak AI: machines can be programmed to exhibit intelligent behavior
• Strong AI: machine can be programmed to possess intelligence and consciousness
• Will AI take over the world? (Elon Musk, Mark Zuckerberg)

Question 4

Understanding images

Answer 4

• Image processing: identifying characteristics of an image:
  
  • Edge enhancement
  • Region finding
• Image analysis: understanding what the image characteristics represent in the real world.

Question 5

Language processing

Answer 5

• Syntactic analysis: identifies the grammatical role of each word (parsing)
• Semantic analysis: identifies the information content in a sentence:
  
  • “Mary gave Johan a birthday card” / “John got a b-day card from Mary”
  • “Do you know what time it is?”
• Contextual analysis: the context is brought into the understanding process:
  “The bat fell to the ground”

Question 6

Production systems

Answer 6

1. Collection of states:
   Start/initial state
   Goal satte/states
2. Collection of productions (rules/moves)
   from one state to another
   might have preconditions
3. Control system: decides what production to apply next

Question 7

Heuristics

Answer 7

• Heuristic: “rule of thumb” for making decisions
• Requirements for good heuristics
  It should constitute a reasonable estimate of proximity to a goal
  It should be easy to compute

Question 8

Learning

Answer 8

• Learning by imitation: The computer records the steps performed by a person
• Supervised learning: a person identifies the correct response for a number of examples and the agent generalizes from those examples
• Learning by reinforcement: the agent is given a general rule to judge for itself when it has succeeded or failed.

Question 9

Artificial neural network

Answer 9

• Artificial neuron:
  each input is multiplied by a weighting factor;
  if sum of weighted inputs exceed threshold then output is 1 else output is 0
• Artificial neural network learn from examples by adjusting their weights

Question 10

Robotics

Answer 10

• Robotics: the study of physical autonomous agents that behave intelligently
• Current robots (only) successful on specific tasks:
  • Mobility: swim like fishes, fly like dragonflies, hop like grasshoppers, crawl like snakes, exploring the surface of Mars
  • Drive a car in traffic
  • Behave like pet dogs
  • Guide weapons to their targets