← Volver al Material

🤖 Technology Class: Artificial Intelligence

🎯 Warm-up Discussion

Let's explore what you already know about AI:

Vocabulary Practice: Key Terms

Before watching the video, review these important AI terms:

Artificial Intelligence (AI): Technology that enables machines to perform tasks requiring human-like intelligence.
Neural Network: A system that mimics how the brain processes information through interconnected nodes.
Machine Learning: The process by which AI systems learn from data without explicit programming.
Generative AI: AI's ability to create new content like text, images, or code.
AI Hallucination: False information generated by AI that sounds credible but is inaccurate.
Model Drift: When AI model performance degrades over time due to changing data patterns.
Training Data: The dataset used to teach an AI model patterns and relationships.
Bias: Unfair prejudice in AI outputs resulting from biased training data.

🎥 Listening Exercise

VIDEO: 5 Things You Really Need to Know About AI

Part A: Technical Details - Multiple Choice

Watch the video and choose the correct answer based on specific information given.

1. According to the video, artificial neural networks originated in:
A) The 1940s
B) The 1950s
C) The 1960s
2. What specific technical process does the video describe for how neural networks improve?
A) They increase processing speed over time
B) They adjust and change their connections based on feedback
C) They add more layers automatically
3. The video states that generative AI applications can:
A) Only generate text responses
B) Create new data
C) Replace traditional databases
4. What does the video say ChatGPT and similar systems actually are?
A) True artificial general intelligence
B) Sophisticated sentence completion applications
C) Advanced rule-based expert systems
5. What technical term does the video use for AI making false statements?
A) AI errors
B) Model drift
C) AI hallucinations
6. According to the video, what fundamental limitation do current AI systems have?
A) Limited processing power
B) Inability to assess truthfulness of their responses
C) Poor user interface design

Part B: Fill in the Blanks - Key Quotes

Listen carefully and complete these important quotes from the video.

1. "A neural network is a bit like a team of interconnected workers that learn to ."
2. "ChatGPT and its counterparts are sophisticated apps."
3. "The core of the technology is a model that uses to predict the next word."
4. "If AI is trained on data that's racist, biased or , then its output will be too."

Reading Exercise: Gapped Text

Read the article about AI development challenges. Six sentences have been removed. Choose which sentence (A-H) fits each gap (1-6). There are two extra sentences you don't need.

Available Sentences:

A. Unlike web applications where response times and error rates are clear indicators, AI systems require monitoring of model accuracy, data drift, and prediction confidence levels.
B. The entire machine learning lifecycle requires careful engineering and infrastructure planning.
C. What works in a research environment often fails when exposed to real-world data volumes and latency requirements.
D. Training large models can cost thousands of dollars, but maintaining them in production environments often exceeds initial development costs.
E. Developers must track data lineage, implement robust validation pipelines, and ensure consistent data formatting across different sources.
F. User interface design for AI applications requires special consideration for uncertainty and explanation of automated decisions.
G. New regulations around AI transparency and accountability mean that developers need to build explainability and auditability into their systems.
H. Traditional software testing approaches are insufficient for validating AI system behavior across all possible inputs.

AI Development: From Concept to Production

Building AI systems that work reliably in production environments presents unique challenges that many developers underestimate. (1) This includes not just the initial training phase, but also ongoing monitoring, retraining, and maintenance of models in live systems.

The data pipeline is often the most critical component. (2) Poor data quality will inevitably lead to poor model performance, regardless of how sophisticated the algorithms are.

Model deployment introduces another layer of complexity. (3) Issues like model versioning, A/B testing frameworks, and rollback strategies become essential when deploying AI systems at scale.

Monitoring AI systems in production differs significantly from traditional software monitoring. (4) Model drift, where the real-world data gradually differs from training data, can cause performance degradation that's difficult to detect without proper metrics.

The computational costs of AI systems often surprise development teams. (5) Inference costs can scale dramatically with user adoption, making it crucial to optimize models for production efficiency.

Finally, the regulatory and ethical landscape continues to evolve. (6) Organizations need to build compliance and ethical review processes into their AI development workflows from the beginning.

Vocabulary Exercise: Match Terms with Definitions

Match each AI term with its correct definition:

1
Artificial Intelligence
2
Neural Network
3
Machine Learning
4
Generative AI
5
AI Hallucination
6
Model Drift

Part B: Technical Context

Choose the correct technical word to complete each sentence:

1. During , raw data is cleaned and formatted before being fed to the model.
2. Model refers to the process of making predictions on new data in production.
3. occurs when a model learns the training data too specifically and fails to generalize.
4. techniques help prevent models from becoming too complex during training.
5. In NLP systems, breaks down text into smaller units for processing.
6. Model involves moving trained models from development to production environments.

💬 Speaking Section: Artificial Intelligence

Discussion cards for meaningful conversation

Do you use AI coding assistants in your work?

💡 Discussion tips:

  • Share tools you've tried: GitHub Copilot, ChatGPT, Claude
  • Discuss benefits vs risks: productivity vs code quality
  • Use vocabulary: machine learning, neural network, training data
Have you experienced AI hallucinations?

💡 Discussion tips:

  • Reference the video concept of AI making up information
  • Share real examples: incorrect code, false information
  • Discuss how to verify AI outputs
How do you ensure training data quality?

💡 Discussion tips:

  • Apply vocabulary: training data, dataset, bias prevention
  • Discuss data validation and cleaning processes
  • Connect to problematic AI outputs from biased data
Should we trust AI with critical decisions?

💡 Discussion tips:

  • Debate healthcare, finance, legal applications
  • Use modal verbs: "AI should/shouldn't", "We must..."
  • Discuss human oversight and accountability
How would you test an AI/ML system?

💡 Discussion tips:

  • Compare to traditional unit testing limitations
  • Discuss model validation, accuracy metrics, edge cases
  • Share experiences with ML testing frameworks
Will AI replace programmers?

💡 Discussion tips:

  • Reference the video's perspective on job evolution
  • Discuss skills that remain valuable: architecture, problem-solving
  • Use future tense: "I think AI will...", "Jobs won't..."
What's the biggest challenge deploying AI models?

💡 Discussion tips:

  • Apply vocabulary: model deployment, inference, production
  • Discuss model drift, monitoring, scaling challenges
  • Share real-world deployment experiences
How do you explain AI limitations to clients?

💡 Discussion tips:

  • Simplify technical concepts: hallucinations, probabilistic outputs
  • Use analogies to make complex ideas understandable
  • Discuss managing client expectations
What's your responsibility for AI bias prevention?

💡 Discussion tips:

  • Discuss ethical considerations in AI development
  • Connect to training data quality and diverse datasets
  • Use modal verbs: "We should...", "Developers must..."
Which AI tools will dominate in 2-3 years?

💡 Discussion tips:

  • Discuss current frameworks: TensorFlow, PyTorch, LangChain
  • Predict future trends: AGI, multimodal AI, agent systems
  • Use future forms: "will become", "are going to replace"

🎯 Conversation Starters:

  • "When the video mentioned hallucinations, I remembered..."
  • "In my experience with AI tools..."
  • "I think the biggest risk of AI is..."
  • "If AI continues developing at this pace..."

💡 Remember to use AI vocabulary: machine learning, neural network, training data, bias, hallucination, model deployment, inference!