Introduction, Access, and Best Practices

This session on 'AI Literacy - Introduction, How to Access, Best Practices' provides a concise overview of Artificial Intelligence, focusing on its relevance and application in academic settings. It includes an introduction to the evolution of AI, an explanation of Prompt Engineering, and the importance of the General Communications Model and Computational Thinking in understanding AI. The session also covers best practices for using AI tools and resources in educational and research environm...

Get Started. It's Free
or sign up with your email address
Introduction, Access, and Best Practices by Mind Map: Introduction, Access, and Best Practices

1. Click and drag to move the mind map

2. Access mind map here https://tiny.cc/ai-literacy-1

3. The need to understand AI

3.1. Alvin Tofller's Future Shock

3.1.1. Speed of technological change exceeds humans ability to adapt

3.2. NotebookLM

3.2.1. https://notebooklm.google.com/

3.2.2. Audience: Faculty Learning Community (FLC) "Advancing AI Literacy @ SacState and Beyond". Style: Professional and grounded. These advancements will affect faculty significantly, it should be engaging but serious. Objective: Explain "Future Shock,” using critical perspectives, in the context of “AI in Higher Education” and “AI and Social Justice” in 2025.

4. Introduction

4.1. The metaphors we use to explain computer processes

4.1.1. Command line constructs

4.1.1.1. Directories Records Cards

4.1.2. "Desktop" as a computer construct

4.1.3. Is it intelligence?

4.2. Large Language Models

4.2.1. What is the most probable word after a sequence of words?

4.2.1.1. Google Search

4.2.1.1.1. what date is the last tv show about

4.2.1.2. Texting in mobile devices

4.2.1.3. Email applications

4.2.1.4. Social media hashtags

4.2.2. How to know what is most probable?

4.2.2.1. Big Data

4.2.2.1.1. Data (a lot of publicaly available text)

4.2.2.1.2. Your data?

4.2.2.1.3. Copyrighted material not on the internet?

4.2.2.2. Fast processing

4.2.2.2.1. A normal CPU runs at 2 Ghz

4.2.2.2.2. New processors run in TeraFlops

4.2.2.2.3. Building an LLM takes a long time

4.3. Brief chronology of AI changes

4.3.1. From completions to interactions

4.3.1.1. Instead of guessing the next word

4.3.1.2. Respond in a conversational manner

4.3.2. Better language models

4.3.2.1. From having "hallucinations" (aberrations?)

4.3.2.2. To reaching ever-higher benchmarks

4.3.2.3. Benchmarks

4.3.2.3.1. https://epoch.ai/data

4.3.2.3.2. Alignment benchmark?

4.3.3. Context

4.3.3.1. Context window size keeps expanding

4.3.3.1.1. 2018 – GPT‑1: ~512 tokens (≈384 words) 2019 – GPT‑2: ~1,024 tokens (≈768 words) 2020 – GPT‑3: ~2,048 tokens (≈1,536 words) 2021 – GPT‑3.5 (ChatGPT): ~4,096 tokens (≈3,072 words) 2023 – GPT‑4 Variants: Standard ~8,192 tokens (≈6,144 words); Extended ~32,768 tokens (≈24,576 words) 2024 – Extended‑Context Models: Meta’s Llama 3.1: ~125,000 tokens (≈93,750 words); Anthropic’s Claude 2.1: ~200,000 tokens (≈150,000 words); GPT‑4 Turbo / GPT‑4o: ~128,000 tokens (≈96,000 words); Google’s Gemini 1.5 Pro: ~2,000,000 tokens (≈1,500,000 words)

4.3.3.2. Three uses of AI (out of many)

4.3.3.2.1. Make a QUESTION to the big data used for a model

4.3.3.2.2. Make a QUESTION about a limited amount of data

4.3.3.2.3. Request a PROCESS based on a limited amount of data

4.3.3.3. Interacting with the context

4.3.3.3.1. Artifacts

4.3.4. Retrieval Augmented Generation (RAG)

4.3.4.1. Pull in data from external sources

4.3.5. "Reasoning" or Chain of “thought”

4.3.5.1. Prompting technique that encourages language models to break down their reasoning into explicit steps, similar to how humans work through complex problems.

4.3.6. "Deep Research"

4.3.6.1. Chain of thought

4.3.6.2. Context

4.3.6.3. Retrieval Augmented Generation

4.4. Multi-modality

4.4.1. Text Processing

4.4.1.1. Input: OCR (Optical Character Recognition), handwriting recognition, document scanning

4.4.1.2. Processing: LLMs, natural language understanding, sentiment analysis

4.4.1.3. Output: Text generation, document creation, translations, summaries

4.4.2. Image Processing

4.4.2.1. Input: Computer vision, object detection, scene understanding, facial recognition

4.4.2.2. Processing: Image segmentation, feature extraction, pattern recognition

4.4.2.3. Output: Image generation (like DALL-E), style transfer, image editing, upscaling

4.4.3. Audio/Voice Processing

4.4.3.1. Input: Speech recognition, voice detection, audio fingerprinting

4.4.3.2. Processing: Natural language processing, tone analysis, acoustic modeling

4.4.3.3. Output: Speech synthesis, voice cloning, text-to-speech

4.4.4. Video Processing

4.4.4.1. Input: Motion detection, frame analysis, action recognition

4.4.4.2. Processing: Temporal analysis, scene segmentation, tracking

4.4.4.3. Output: Video generation, animation synthesis, motion transfer

4.4.5. 3D/Spatial Processing

4.4.5.1. Input: 3D scanning, depth sensing, LIDAR data

4.4.5.2. Processing: Point cloud processing, mesh reconstruction, spatial mapping

4.4.5.3. Output: 3D modeling, virtual environment generation, augmented reality overlays

4.4.6. Benchmark vs Human Performance

5. Accessing AI

5.1. Privacy first

5.1.1. Google Account

5.2. Google AI Studio

5.2.1. https://aistudio.google.com/prompts/new_chat

6. Best practices

6.1. Prompts

6.1.1. Information that will trigger the probabilistic processes

6.2. Prompt engineering

6.2.1. Proficiency at controlling the probabilities

6.3. Parts of a prompt

6.3.1. Core Task Subject Matter Scope Role Assignment Tone Format Length Components

6.3.1.1. Representation of tokens in different queries

6.3.2. https://docs.google.com/spreadsheets/d/159vc98MMU-XbD_kO0tuaaVjndjR6Cj1bU1_K0kplCUY/edit?usp=sharing

6.3.3. Provide examples Provide a series of steps Split complex tasks into smaller tasks

6.4. Controlling the temperature

6.4.1. Representation of the temperature

7. Activities

7.1. Interact with an LLM and explore using the parts of a prompt

7.1.1. Claude AI

7.1.1.1. Sonnet 3.5

7.1.2. ChatGPT

7.1.2.1. o3-mini, o1, or 4o

7.1.3. Google AI Studio

7.1.3.1. Gemini 2

7.1.3.1.1. https://aistudio.google.com/prompts/new_chat

7.2. Data Literacy

7.2.1. https://datadetoxkit.org/en/home

7.3. Interact with Google Labs FX

7.3.1. https://labs.google/fx