Click to edit Master title style

Types of
knowledge
By- Rahul Bingumalla

1
Click to edit Master title style

Declarative Meta Structural

Knowledge Knowledge Knowledge

Procedural Heuristic
Knowledge Knowledge

2
2
Click to edit Master title style

Declarative Knowledge

Subtitle

3 3
Click to editKnowledge
Declarative Master title style

• What is Declarative Knowledge?

• It's the knowledge of facts and information that we can consciously
recall and express.
• It's often called "knowing what" instead of "knowing how."
• It encompasses various topics, from simple facts to complex
concepts.
• It's stored in our declarative memory, which is distinct from our
procedural memory (which stores knowledge of how to do things).

4 4
Click to edit Master title style

• Examples of declarative knowledge:

• A language model knowing the definition of every word in its
vocabulary.
• A computer vision system recognising different objects and their
properties (a cat is furry, has four legs, and meows).
• A chess AI understands the game’s rules and the possible moves for
each piece.
• A medical diagnosis system knowing the symptoms and
characteristics of various diseases.

5 5
Click to edit Master title style

Procedural Knowledge

Subtitle

6 6
Click to editKnowledge
Procedural Master title style

• What is Procedural Knowledge?

• It's the knowledge of how to do things, encompassing the
steps, processes, and rules involved in performing tasks.
• It's often called "knowing how" as opposed to "knowing what."
• In AI, it's crucial for enabling systems to make decisions, solve
problems, and take action autonomously.

7 7
Click to edit Master title style

• Examples:
• [Imagine a flowchart] A chatbot following a decision tree to guide a
customer through troubleshooting steps.
• [Visualize a robot arm] A robot assembling a product by following a
sequence of precise movements.
• [Picture a self-driving car] A self-driving car navigating traffic, making
decisions based on its understanding of the road and traffic rules.
• [Imagine a machine translation algorithm] A machine translation
system generates text in another language by following grammatical
rules and patterns.
• [Visualize a game-playing AI] A chess-playing AI selects moves based
on its knowledge of strategies and tactics.

8 8
Click to edit Master title style

Meta Knowledge
Subtitle

9 9
Click
Meta to edit Master title style
Knowledge

•
Meta-knowledge in AI takes things further than simply knowing facts
or performing tasks. It's about knowing what you know, understanding
the limitations of your knowledge, and being able to reason and learn
about different ways of acquiring and using knowledge. Here's how it
works in AI, with some real-life examples:

1010
Click to edit Master title style

• Examples:
• Imagine a self-driving car:
• Regular knowledge: Traffic rules, map data, object recognition.
• Meta-knowledge: Knowing its sensor limitations (e.g., blind
spots), understanding the uncertainty in its predictions, and
recognizing situations where external help might be needed
(e.g., extreme weather).
• Think of a chess-playing AI:
• Regular knowledge: Opening strategies, endgame tactics, evaluating
positions.
• Meta-knowledge: Knowing its strengths and weaknesses against
different opponent styles, adapting its playing style based on the
game situation, and learning from its mistakes to improve future
strategies.

1111
Click to edit Master title style

Heuristic Knowledge
Subtitle

1212
Click to edit
Heuristic Master title style
Knowledge

• Heuristic Knowledge refers to imprecise but often helpful rules of

thumb that guide decision-making and problem-solving. It's like having
a seasoned expert whisper advice in your ear, helping you navigate
complex situations without complete information.
• Think of it as shortcuts or mental frameworks that approximate
optimal solutions when precise calculations or exhaustive analysis are
impractical. These shortcuts are often learned through experience or
observation, making them well-suited for real-world scenarios where
data might be incomplete or constantly changing.

1313
Click to edit Master title style

• Search Algorithms:
• Heuristics like "move towards the nearest food source" guide robots
navigating mazes or foraging for resources. These simple rules, while
not guaranteeing the shortest path, help the robot make progress
efficiently.
• 4. Spam Filtering:
• Email filters often use heuristics like identifying keywords or sender
patterns to flag potential spam. These shortcuts help keep inboxes
clean while acknowledging that some legitimate emails might get
caught in the net.

1414
Click to edit Master title style

Structural Knowledge
Subtitle

1515
Click to edit
Structural Master title style
Knowledge

• refers to the understanding of how concepts and information are

related to each other. It acts as the glue that binds together the
building blocks of knowledge, giving AI systems the ability to reason,
infer, and solve problems effectively.
• Think of it like building a house: you need individual bricks (facts and
concepts) and a blueprint (structural knowledge) to put them together
in a meaningful way. Without the blueprint, the bricks remain a
scattered pile, lacking the structural integrity to stand as a house.

1616
Click to edit Master title style

• Examples:
• Recommender systems: By understanding user preferences and the
relationships between products, these systems recommend items
likely to be interesting.
• Fraud detection: AI systems analyse financial transactions and
identify patterns indicative of fraudulent activity using knowledge of
typical transaction structures and relationships.
• Natural language processing: AI systems understand the grammatical
and semantic structure of language, allowing them to parse
sentences, translate languages, and generate text.

1717
Click to edit Master title style

Conclusion:
In artificial intelligence, knowledge representation is crucial in enabling
machines to comprehend and manipulate information effectively.
Different types of knowledge serve distinct purposes, and
understanding their nuances is essential for designing intelligent
systems.
Effective AI systems often combine multiple knowledge representation
types to achieve complex goals. By carefully selecting and integrating
these representations, we can create machines that learn, reason, and
solve problems with increasing sophistication, transforming diverse
fields and shaping our future.

18
Click to edit Master title style

19