Aplicaciones de la Inteligencia Artificial - Video Explicativo (Fundamentos de la IA)

Main ideas, concepts, and lessons

• Purpose of the module: Introduce what Artificial Intelligence (AI) is, how it emerged and evolved, and the main types/levels of AI.

Definition of AI (Royal Spanish Academy - RAE)

• AI is a scientific discipline focused on computer programs that perform operations similar to those associated with human intelligence (e.g., learning or logical reasoning).
• The idea is to imitate cognitive capacities such as:
  • learning from experience,
  • solving problems,
  • making decisions.
• The speaker emphasizes that AI has multiple definitions, depending on goals, methods, and criteria.

Other notable conceptions of AI (examples from authors)

• AI as making computers do tasks humans do better (human-comparison framing).
• AI as science/engineering of intelligent machines, especially intelligent computer programs (Luger & Knight).
• AI as a branch of computer science focused on automating intelligent behavior (McCarthy).

---

Factors influencing AI development (historical timeline)

1950s–1960s

• Development of programming languages and algorithms for complex problem solving.
• Established tools/methods so computers could handle tasks requiring human-like intelligence (reasoning, learning, decisions).

1970s–1980s

• Emergence of neural networks and expert systems.
• Presented as major AI branches/paradigms based on different ways of representing and processing knowledge.

1990s–early 21st century

• Advances in deep learning and robotics.
• Emphasis on innovative machine learning techniques and machines that can interact with the physical environment.

---

Stages (phases) of Artificial Intelligence (4 main phases)

Phase 1: Rise of Artificial Intelligence

• Early programs solving problems in:
  • Logic
  • chess
  • geometry
  • algebra
• Early expert systems emulating human reasoning within specific domains.
• Examples:
  • DENDRAL: analyzed chemical compounds
  • MYCIN: diagnosed infections

Phase 2: Winter of Artificial Intelligence

• Crisis/stagnation due to technical and theoretical limitations.
• Problems mentioned:
  • difficulty representing common knowledge
  • challenges with combinatorial equations
  • lack of resources and funding
  • loss of credibility

Phase 3: Renaissance of Artificial Intelligence

• Recovery and progress due to new technologies and methodologies.
• Contributing factors:
  • development of artificial neural networks and machine learning
  • rise of the internet and distributed systems
  • growth of new domains and applications

Phase 4: Artificial Intelligence Revolution

• Unprecedented expansion driven by exponential advances in technology and data.
• Milestones listed:
  • IBM Watson winning Jeopardy!
  • Google DeepMind winning the game of Go
  • development of the autonomous car

---

Types and levels of Artificial Intelligence

1) Weak vs. strong AI (based on relationship to humans)

• Weak / Narrow AI (currently existing)

  • Performs specific, limited tasks.
  • Does not have awareness or understanding of what it is doing.
  • Framed as the only type existing today.

• Strong / General AI (hypothetical ideal)

  • Could perform any task a human can do with awareness and understanding.
  • Described as hypothetical; does not exist currently.

• Superintelligent AI (speculative/futuristic)

  • Would far surpass human capabilities in all aspects.
  • Presented as speculative.

2) Reactive vs. memory/goal-based (based on autonomy and interaction)

• Reactive AI

  • Only responds to external stimuli, with no memory or learning.
  • Example: Deep Blue (chess).

• Limited AI

  • Has short-term memory.
  • Can learn from past experiences to improve performance.
  • Example: AlphaGo (Go).

• Theoretical AI (as described)

  • Would have long-term memory.
  • Could reason about its own goals and plans.
  • Example (fiction): HAL 9000 from 2001: A Space Odyssey.

• Self-aware AI (as described)

  • Would be conscious of itself.
  • Could interact with other intelligent agents.
  • Example (fiction): Data robot from Star Trek.

Note as stated in subtitles: the speaker mentions “there is no such thing as reactive artificial intelligence,” but the remainder of the video immediately defines reactive AI and gives an example—so the subtitles contain a likely inconsistency.

3) Horizontal vs. vertical AI (based on scope/domain/application)

• Horizontal AI

  • General-purpose capability (example: speech recognition).
  • Speech recognition systems:
    • use signal processing algorithms + machine learning
    • convert human speech into text
    • can be trained to recognize many languages and accents
    • do not require deep knowledge of a specific domain

• Vertical AI

  • Specialized for a specific field/domain.
  • Example: medical diagnosis.
  • Medical diagnosis systems:
    • use ML and data processing to analyze:
      • MRI images
      • laboratory test results
      • medical records
    • are trained to detect patterns and signals indicating diseases/conditions
    • intended to support clinicians with more accurate and efficient diagnoses

---

Speakers / sources featured (as named in the subtitles)

• Royal Spanish Academy (RAE) (source of the definition)
• Luger (named as “Luger … Kevin Knight”; exact first name unclear due to subtitle errors)
• Kevin Knight
• John McCarthy
• DENDRAL (expert system example; referenced by name)
• MYCIN (expert system example; referenced by name)
• IBM Watson (milestone example; referenced by name)
• Google DeepMind (milestone example; referenced by name)
• Deep Blue (chess example)
• AlphaGo (Go example)
• HAL 9000 (2001: A Space Odyssey) (fictional example)
• Data (Star Trek) (fictional example)
• Computer/science institutions/products/tools (not explicitly persons): MRI, autonomous car (referenced as technologies/examples)