Gestures Off

Introduction to Artificial Intelligence

Preparing for the AI Future: Awareness, Capabilities, and Responsible AI Practice[cite: 2]

I. Key AI Concepts

Artificial Intelligence (AI): A set of technologies that enable computers to reason, learn, and act in ways that would typically require human intelligence[cite: 2]. Defined as "the science of making machines do things that would require intelligence if done by men," AI is the first technology capable of independent decision-making[cite: 2].

Machine Learning (ML): A subset of AI focused on enabling systems to learn from data and improve performance on tasks without explicit programming[cite: 2].

Deep Learning (DL): A subset of ML utilizing neural networks to achieve advanced learning capabilities, particularly with complex, unstructured data[cite: 2]. Most Large Language Models (LLMs) rely on Deep Learning[cite: 2].

Generative AI (Gen AI): AI used to generate new material—such as text, images, music, audio, and videos—by identifying patterns and correlations from training data[cite: 2].

Large Language Models (LLMs): A subset of Gen AI specifically built to understand and generate human language through unsupervised/self-supervised learning on massive text corpora[cite: 2].

Prompt Engineering: The practice of designing precise instructions, context, and constraints to guide an AI system toward desired, high-quality results[cite: 2].

AI Hallucinations: Convincing but incorrect or fabricated outputs generated by AI models due to training gaps, biased data, or faulty inferences[cite: 2].

II. Prompt Engineering Best Practices

  1. Put instructions at the beginning: State the core task before providing context or data[cite: 2].
  2. Use delimiters: Employ quotation marks, triple backticks, or XML tags to cleanly separate instructions from referenced input[cite: 2].
  3. Be very specific: Define target output formats, intended audience, length, and constraints[cite: 2].
  4. Assign a persona: Give the model a role or perspective to shape tone and depth[cite: 2].
  5. Provide relevant examples: Include few-shot input/output examples to steer formatting and voice[cite: 2].
  6. Ask for a chain of thought: Prompt the model to work through steps systematically before finalizing answers[cite: 2].
  7. Specify the output format: Mandate structured formats such as Markdown tables, bullet lists, or code blocks[cite: 2].
  8. Supply grounding data: Insert factual data directly into the prompt rather than assuming internal recall[cite: 2].
  9. Ask for evidence: Instruct the model to cite sources or explicitly state "I don't know" when unsure[cite: 2].
  10. Prompt in iterations: Start simple, evaluate results, and iteratively refine constraints[cite: 2].

III. AI Tools & Use Cases

Generative AI streamlines drafts, translations, summarization, research workflows, and coding[cite: 2]. Specialized tools highlighted include:

Perplexity AI

Focus: Factual conversational search engine[cite: 2].

Key Feature: Generates real-time, sourced answers with auto-cited links back to authoritative data[cite: 2].

Google NotebookLM

Focus: Document-grounded research assistant[cite: 2].

Key Feature: Synthesizes notes, PDFs, and docs; grounds answers exclusively in your uploaded source material[cite: 2].

Google Pinpoint

Focus: Large-scale investigative analysis[cite: 2].

Key Feature: Transcribes multi-language audio, extracts entities, and searches across thousands of unstructured files[cite: 2].

IV. Ethics & Risk Management Framework

  • Human-in-the-Loop: Retain mandatory human oversight and verification over critical decisions[cite: 2].
  • Bias Mitigation & Fairness: Audit datasets to ensure representative sampling and avoid discrimination based on demographic factors[cite: 2].
  • Data Privacy & Security: Enforce strict safety protocols to prevent leaks and comply with privacy legislation[cite: 2].
  • Transparency & Explainability: Ensure AI behavior and reasoning remain interpretable to end users[cite: 2].
  • Accountability: Establish legal responsibility, operational safety policies, and ongoing risk monitoring before deployment[cite: 2].