Symbolic AI

#SymbolicAI, often referred to as Good Old-Fashioned Artificial Intelligence (GOFAI), dominated the early stages of AI research. It’s rooted in the idea that intelligence can be replicated by manipulating symbols based on explicit rules and logic. In essence, it aimed to encode human knowledge and reasoning directly into machines.

Key techniques within Symbolic AI include methods such methods as

• #RuleBasedSystems: These systems use ”if-then” rules to make decisions. For example, ”IF temperature is high AND pressure is low THEN open valve.”
• #FuzzyLogic: This approach handles uncertainty by allowing degrees of truth rather than strict true/false values. For instance, ”IF temperature is somewhat high THEN slightly increase cooling."
• #ExpertSystems: These systems mimic the decision-making of human experts in specific domains, using a knowledge base and inference engine. Symbolic AI achieved notable successes in areas like game playing (e.g., Deep Blue beating Garry Kasparov) and limited expert systems. However, it struggled with tasks requiring perception, learning from raw data, and handling complex, real-world environments.

However in the 1990s, due to the brittleness, difficulty in scaling, and inability to learn from data, led to its decline in favour of statistical based AI, such as #MachineLearning. Where the focuses on learning patterns from data through interconnected network of nodes, such as Neural networks. These excelled in tasks that were challenging for symbolic AI, such as image recognition, speech processing, and natural language understanding.

Symbolic AI, while facing limitations in the past, is experiencing a resurgence through #NeuroSymbolicAI, aka #IntegratedAI. Additionally, it remains relevant in areas like #PedestrianEvacuation modelling, where rule-based systems provide valuable insights. This hybrid approach leverages the strengths of both symbolic reasoning and statistical AI (Machine Learning), addressing critical challenges in modern AI, such as control, explainability, generalisation, bias and reducing hallucinations. As AI continues to evolve, the integration of these paradigms will play a crucial role in building more intelligent and reliable systems.