Symbolic AI

As explained in Build a Reasoning Model (From Scratch)

Reasoning based on explicit symbolic rules rather than statistical associations learned from data.

Rule-based reasoning uses explicit symbolic rules and logical inference rather than statistical associations learned from data. It provides a reference point for understanding how LLM reasoning differs: LLMs can generate intermediate steps and simulate reasoning behavior without explicitly executing rule-based logic. section 1.5 section 1.8

Use the distinction to separate the mechanisms and improvement methods involved:

• Rule-based systems apply explicit logical rules; LLMs rely on statistical pattern matching learned from data. section 1.8
• LLM reasoning can be improved through inference-time compute scaling, reinforcement learning, supervised fine-tuning, and distillation without becoming rule-based. section 1.5 section 1.8

Figure 1.7 An illustrative example of how a language model (GPT-4o in ChatGPT) appears to “reason” about a contradictory premise

Inspect an example of apparent LLM reasoning about a contradictory premise, then compare its behavior with explicit rule execution. section 1.5

How Build a Reasoning Model (From Scratch) covers it

How deep it goes

Defines the distinction and compares mechanisms, but stops short of specifying a rule-engine implementation or inference algorithm.

Examples

Uses an LLM response to a contradictory premise as an example of apparent reasoning without explicit rules.

What you can do after

Evaluate whether a reasoning behavior reflects explicit logical inference or statistical pattern matching.

How it connects

Connects rule-based reasoning with logical reasoning, statistical pattern matching, chain-of-thought, reinforcement learning, and inference-time compute scaling.

liveBookmanning.com

Read Build a Reasoning Model (From Scratch) on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  The Art of AI Product DevelopmentA rule-based approach to artificial intelligence derives behavior or conclusions from explicitly specified rules rather than patterns learned from data.
• 

  liveBookmanning.com
  Deep Learning with R, Third EditionSymbolic AI is an artificial-intelligence approach that represents knowledge explicitly and manipulates it with hand-crafted symbolic rules.
• 

  liveBookmanning.com
  Deep Learning with Python, Third EditionAn AI approach that represents knowledge and reasoning using explicit symbols and rules rather than learned representations.
• 

  liveBookmanning.com
  Inside AISymbolic AI is an artificial-intelligence approach that represents knowledge explicitly with symbols, rules, and logic.

Compare how 5 books explain symbolic AI →

Related concepts

• machine learningalternative to · Machine learning is a field of computational methods in which systems learn patterns or relationships from data to make predictions, decisions, or other outputs.
• regular expressionuses · A formal pattern notation that describes sets of strings and enables text to be matched, searched, extracted, or transformed.
• rule of logicuses · A rule of logic is a formal conditional statement specifying how conclusions follow from given facts or conditions.
• artificial intelligencekind of · Artificial intelligence is the field concerned with creating computational systems that perform tasks associated with human intelligence, such as learning, reasoning, perception, or decision-making.

This explainer was generated by Manning's knowledge graph from the text of Build a Reasoning Model (From Scratch). The section links point to the parts of the book it draws on.