Complete Artificial Intelligence Study Guide

Detailed Review – AI Concepts, LLMs, Risks, Ethics, and Business Impact

I. Core Definitions

• Artificial Intelligence: Software that mimics or improves tasks requiring human intelligence.
• Generative AI: AI that creates text, images, code, audio, or video from prompts.
• LLM: Large Language Model trained on massive datasets.

II. AI Before ChatGPT

• Google searches
• Ads
• Spam filters
• Spell checkers
• Recommendation systems
• Data mining
• Mainstream consumer technology

III. What Happened in 2022?

• ChatGPT 3.5 launched publicly.
• Generative AI became mainstream.
• Large jump in public awareness and adoption.

IV. Hallucinations

• False answers stated confidently.
• Can invent facts, sources, and details.
• Requires verification.

V. Bias

• Can reflect racial, gender, and social bias from data.
• Examples include hiring systems and image generation.

VI. How Language Models Work

• Learn patterns in language.
• Represent meaning mathematically.
• Predict likely next words/tokens.

VII. Transformers

• Generative Pre-Trained Transformer.
• Analyzes context efficiently.
• Enables advanced modern AI.

VIII. Training Terms

• Corpus: Data used for training.
• Parameters: Internal learned values.
• More compute and data usually improve performance.

IX. Best Uses

• Summaries
• Brainstorming
• Draft writing
• Learning new topics
• Quality improvement

X. Risks

• Copyright
• Security
• Privacy
• Misinformation
• Job shocks

XI. Security Risks

• Data ingestion leaks
• Prompt injection
• Data poisoning

XII. Resource Costs

• Millions of dollars to train large models.
• High electricity and water use.

XIII. Ethical Issues

• Who is responsible for decisions?
• How should autonomous systems behave?
• Need for transparency and fairness.

XIV. What Organizations Can Do

• Hire diverse teams
• Hire ethicists
• Create ethics policies
• Use review boards
• Audit systems
• Train employees

XV. Red Teaming

• Independent testers try to break the system before release.
• Find weaknesses and improve safety.