Inteligencia artificial en el aula con Scratch 3.0 - Presentación del Tutorial

Main ideas, concepts, and lessons

• Purpose of the video/resource (CodeINTEF):

  • Introduce Artificial Intelligence (AI) to Primary and Secondary students in a way that is simple, practical, and fun.
  • Promote classroom learning so future decision-makers understand both capabilities and limitations of AI systems.

• What AI is (basic definition):

  • AI is the science/engineering of building systems that perform tasks requiring human-like intelligence or reasoning.

• Why AI has advanced recently:

  • AI is not entirely new—humans have long tried to automate intellectual tasks (even before early computers).
  • Recent achievements are driven by:
    • Better environments where AI algorithms run
    • Large amounts of available data

• Examples of AI success:

  • Autonomous vehicles
  • Games/competitions against human champions such as:
    • Chess
    • Jeopardy
    • Go

• Focus area: Machine Learning

  • Machine Learning is presented as a subset of AI.
  • Key shift from classical programming:
    • In traditional programming, developers write rules by hand
    • In machine learning, the computer learns rules automatically from many examples
  • Thus, the system is trained rather than directly programmed.

Practical example to explain machine learning

• Goal:

  • Build a program to classify books as crime novels vs. other genres using book descriptions.

• Manual rule approach (problematic):

  • Example rule: If a description contains “policeman”, label it as a crime novel.
  • Issue: It misclassifies books where “policeman” appears but the book is not a crime novel (e.g., a romance with a policeman).

• Refined rule (still insufficient):

  • Example: “policeman” + “crime”
  • Issue: Still fails for descriptions where police are absent because crimes are no longer committed (alternate-universe fiction).

• Why this is hard manually:

  • Defining the correct set of rules is difficult, time-consuming, and costly.

Machine learning solution (methodology/instructions)

• A step-by-step approach is described:

  1. Collect training data
     • Take a number of book descriptions and split them into two groups:
       • Crime novels
       • Other types
  2. Train a model
     • Use these examples so the system learns patterns that distinguish crime-novel descriptions.
  3. Classify new items
     • Apply the learned model to classify new book descriptions not used during training.
  4. Iterate/scale
     • The system can still make mistakes, but using more training examples increases the likelihood of correct classification.

Where machine learning is used

The video lists common real-world applications:

• Spam filters
• Music/movie recommendation systems
• Translation services
• Search engines
• Fraud detection systems

Why teach it in schools

Since AI-driven tools will appear more in everyday life, people in all professions need to understand:

• How AI systems work
• How to use them appropriately
• Their drawbacks and limitations

Making machine learning accessible to students

• Historically, machine learning education was mainly for later university-level students in computer engineering/telecommunications.
• Now it’s possible for earlier education using new graphical tools.
• The resource demonstrates how to use Machine Learning for Kids, which:
  • Uses IBM Watson Developer Cloud APIs
  • Lets users integrate AI features into Scratch 3.0 via blocks
• Intended outcome:
  • From the last years of primary school through secondary (academic/vocational), teachers can create Scratch projects so students can classify:
    • Texts
    • Numbers
    • Images

Speakers / sources featured (from the subtitles)

• CodeINTEF (mentioned as the provider of the monthly resource)
• IBM Watson Developer Cloud APIs (technology mentioned as used by “Machine Learning for Kids”)
• Scratch 3.0 / Scratch (platform referenced for implementing projects)
• Machine Learning for Kids (platform referenced as the instructional tool)