1.4: Artificial Intelligence

1.4.1. Artificial Intelligence definition

1.4.2. How does AI works

Here is a quick breakdown of how AI actually ingest data to build a model:

• Supervised Learning (Learning with a Teacher):
  • Concept: The model is trained on labeled data (input data paired with the correct answer).
  • Example: Feeding a computer 10,000 photos explicitly labeled "Cat" or "Not Cat" until it learns the distinct features of a cat.
• Unsupervised Learning (Finding Hidden Patterns):
  • Concept: The model is given unlabeled data and must find inherent patterns, structures, or groupings on its own.
  • Example: Customer segmentation in marketing—grouping millions of buyers into distinct personas based solely on buying behavior.
• Reinforcement Learning (Trial and Error):
  • Concept: An agent learns to make decisions in an environment by performing actions and receiving rewards or penalties.
  • Example: Teaching an AI to play a video game or navigate a robotic arm by rewarding it for getting closer to the goal.

1.4.3. The AI hierarchy

1.4.4. Real-World Applications & Current Impact

Let's see how these subfields manifest in day-to-day industries:

• Healthcare: Computer vision algorithms analyzing X-rays and MRIs to detect anomalies faster than the human eye.
• Finance: Fraud detection systems scanning millions of credit card transactions per second to catch anomalies.
• Transportation: Autonomous vehicles utilizing sensor fusion (combining cameras, radar, and lidar data) to navigate traffic.
• Generative AI: Large Language Models (LLMs) and transformer architectures processing natural language to assist in coding, drafting text, and brainstorming.

Recommended movie

Her is a 2013 American science fiction romantic drama film written, co-produced, and directed by Spike Jonze. 

Starring: Joaquin Phoenix, Amy Adams, Rooney Mara, Olivia Wilde, Scarlett Johansson.

1.4.5. The Critical Wrap-Up: Ethics & Limitations

Remember that AI isn't magic—it is math running on infrastructure, which comes with distinct human responsibilities:

• Algorithmic Bias: If an AI model is trained on historical data that contains human bias, the AI will perfectly reproduce and scale that bias.
• Hallucinations: Generative models predict the next most probable word or pixel; they do not possess a true "fact-checking" compass, meaning they can confidently assert falsehoods.
• The "Black Box" Problem: Deep learning models are incredibly complex. Often, engineers know what output a model produced, but it is incredibly difficult to map out the exact internal pathways of why it made that decision.

1.4.6. Generative AI

Instead of simply organizing existing information, it predicts what element should come next to build something entirely new.