What You Will Be Able To Do

• Describe how AI changes tasks in jobs.
• Set personal rules for AI in study.
• Summarize copyright questions.
• Explain the EU AI Act tiers.
• Apply a responsible-use checklist.

Recap Of Week 10

• API calls need model, messages, max tokens.
• Chat apps resend history each turn.
• Validate JSON; retry on failure.
• Track tokens for cost; stream for speed.

Work: Think In Tasks

• A job is a bundle of tasks.
• AI speeds up some tasks, others stay human.
• Estimates of exposure are uncertain.
• Technology changes roles, and transitions are hard.

Skills That Gain Value

• Asking clear questions.
• Checking and judging outputs.
• Deep knowledge of a field.
• Working with people and taking responsibility.

Education: Tutor Or Crutch

• Explains, quizzes, and gives feedback any hour.
• Effort builds memory; skipping it costs learning.
• Detection tools are unreliable.
• Process-based assessment is growing.

A Personal Study Rule

• Try the problem yourself first.
• Then ask AI to check, explain, or challenge.
• Follow your institution's policy.
• Say openly when you used AI.

Law: Who Is Responsible

• People and organizations stay responsible.
• Existing laws already apply to AI uses.
• Lawyers have been sanctioned for invented citations.
• Rules differ by country; this is not legal advice.

Copyright: Two Questions

• Is training on copyrighted works allowed?
• Courts in several countries are deciding.
• Who owns AI output?
• US view: human authorship is needed for protection.

Copyright: Practical Habits

• Check licenses of inputs and outputs.
• Do not pass off AI text as your own research.
• Record how you used AI in creative work.
• Expect rules to keep changing.

The EU AI Act: The Idea

• Regulation (EU) 2024/1689, in force since August 2024.
• Obligations phase in over several years.
• Rule by risk: more harm potential, stricter rules.
• Applies to providers and users serving the EU.

The EU AI Act: Four Tiers

• Unacceptable: banned, such as social scoring.
• High: hiring, education, credit, law enforcement.
• Limited: tell people they are talking to AI.
• Minimal: most other systems.

banned   -> social scoring by authorities
high     -> hiring, credit, exams
            risk mgmt, data, oversight
limited  -> disclose AI, label content
minimal  -> no extra duties

General-Purpose Models

• Providers of large models have their own duties.
• Documentation for downstream developers.
• A copyright policy.
• Extra duties for the most capable models.

Beyond Europe

• US: existing laws, agency guidance, state laws.
• China: specific rules for generative AI.
• Voluntary frameworks: NIST AI RMF, OECD principles.
• Serving many places: follow the strictest rule.

Responsible-Use Checklist

• Purpose, people, transparency, oversight.
• Data, accountability, environment.
• Score each: strong, partial, weak.
• Fix the weakest first.

1 Purpose        2 People harmed?
3 Transparency   4 Human oversight
5 Data rights    6 Accountability
7 Smallest adequate model

Recap

• Think of work in tasks; checking skill gains value.
• Use AI to learn, not to skip learning.
• People stay legally responsible.
• Copyright questions remain open; keep records.
• The EU AI Act regulates by risk tier.

Next Week: Capstone

• Design, build, and present a small AI assistant.
• Brief, rubric, demo script, review checklist.

Questions And Exercise

• Review one real AI use with the checklist.
• Place it in an AI Act tier and justify.
• Interview someone from another field.