Introduction to AI 2

by Albert Major

Context

• Topic: Rodney Brooks’ critique of traditional AI and his alternative approach (subsumption architecture, embodied robotics).
• Focus: What Brooks criticized in 1990s AI, his new ideas, and their impact.
• Main Reading: “Intelligence Without Representation” by Rodney Brooks.

This slide lists the three main **discussion questions** for the lecture on Rodney Brooks’ “Intelligence Without Representation.” It asks what **aspects of AI** Brooks criticized in the 1990s (such as the focus on **representation** and **centralized control**), to **summarize his radical new approach** (subsumption architecture, embodied robotics), and to consider if Brooks’ approach is **still relevant** today. These questions frame the lecture’s focus on **Brooks’ critique**, his **contribution**, and the **impact** of his ideas on modern AI.

1. Why This Paper Is Important

• Landmark in robotics and AI development.
• Introduced a radically different view from most AI work at the time.
• Challenging to read because of specialized terms, but foundational for understanding robotics today.

This slide summarizes Brooks’ main criticisms of classical AI discussed in the lecture. It lists how Brooks challenged the focus on human-like intelligence, the assumption of a central system (like a brain), the fragmentation into isolated subfields (e.g., planning, representation), and the engineering mindset in AI research. It also explains why AI became obsessed with representations—to link these isolated parts. The slide is important because it highlights exactly what Brooks saw as the core problems in AI’s traditional approach and sets up why his alternative was needed

2. Brooks’ Critique of Classical AI

• Classical (Traditional) AI Approach:
  • AI focused on representation: turning real-world problems into symbols and data structures the computer can work with.
  • Problems solved in “micro-worlds” (e.g., chess, block world), not the real physical world.
  • AI tried to copy human intelligence, especially “thinking” tasks like planning and problem-solving.
  • Assumed intelligence meant having a “central brain” (CPU) doing all the thinking and control.
• Key Problems Brooks Identified:
  • Overemphasis on Human Intelligence: AI tried too hard to mimic the way humans think (e.g., chess), ignoring that most intelligent life forms (like insects) behave very differently.
  • Centralized Control: AI assumed intelligence required a central system (like a brain or a computer program) that takes in information (sense), thinks (plan), then acts.
  • Fragmented Subfields: AI split intelligence into separate modules (e.g., learning, planning, knowledge) that don’t connect well.
  • Representation Obsession: AI relied heavily on making models of the world (representations), which often missed the real-world’s complexity and dynamism.
  • Neglect of Physical Embodiment: Most AI ignored the importance of having a body or acting in the real world.
  • AI as Engineering, Not Science: AI researchers sometimes acted like engineers who try to build planes without understanding aerodynamics—focused on building stuff rather than understanding the deeper principles.

This slide presents a quiz question about Brooks’ main criticisms of AI in the 1990s. It asks which statements are correct: Brooks believed AI focused too much on specific aspects (correct), and that AI’s obsession with representations was wrongheaded (also correct). He did not think AI should slow down, nor that it should focus on high-level cognition first. This slide is important because it checks students’ understanding of Brooks’ real positions, reinforcing his critiques of representation and fragmented focus in classical AII

This slide asks students to summarize Brooks’ radical new approach to AI. The key is his move away from internal representations toward embodied robotics and the subsumption architecture—using simple, layered controllers that directly link sensing to action, without a central “boss” or internal model. This slide is important because it focuses on understanding Brooks’ main contribution and why it was so different from previous AI approaches.

3. What Is Representation in AI?