AI and Generative AI – Uses and Applications in Education

judyl@cuhk.edu.hk

21 November 2024

Questions

1

What comes to mind when you think about AI?

2

Examples or applications of AI in daily life, teaching, or learning experiences?

Microsoft Copilot for the Web

Microsoft Copilot: Your everyday AI companion

Microsoft Copilot: Your everyday AI companion

Microsoft Copilot leverages the power of AI to boost productivity, unlock creativity, and helps you understand information better with a simple chat experience.

AI Definition

Simulation

AI simulates human intelligence processes.

Machine Performance

Machines perform tasks requiring human intelligence.

Key Abilities

Learning, problem-solving, and decision-making.

AI History

1

1950s

Alan Turing and John McCarthy pioneer AI concepts.

2

Turing Test

Proposed in 1950 to evaluate machine intelligence.

3

Ongoing Impact

Turing Test remains a cornerstone in AI evaluation.

digitalwellbeing.org

Artificial Intelligence Timeline Infographic – From Eliza to Tay and beyond

Do you know your Eliza from your Tay, and the other A.I. heroes and villains from sixty years of Artificial Intelligence? Well, here’s an updated and downloadable A.I. timeline infographic to…

AI Evolution Phases

1

Rule-based Programming

Early AI systems followed predefined rules.

2

Machine Learning

AI learns from data with minimal human intervention.

3

Neural Networks

Complex systems mimicking human brain function.

4

Generative AI

Creates new, original content based on learned data.

Enterprise AI

What is Gen AI? Generative AI Explained | TechTarget

Generative AI is a type of artificial intelligence technology that can produce various types of content. Find out how it works and why it's a hot commodity.

Specialized to Generalized AI

1

Specialized Tasks

Early AI focused on narrow tasks like playing chess or recognizing patterns in images.

2

Adaptive Learning

AI evolved to learn from data and adapt to new situations, making it more flexible.

3

Generalized Use

The latest AI systems can perform a wide range of tasks, making them more applicable across different fields.

Large Language Models (LLMs)

GPT-3

OpenAI's powerful language model.

ChatGPT

Conversational AI capable of human-like text generation.

Other Models

Claude by Anthropic and LLaMA 3 by Meta.

ChatGPT (2022): 1 millon users in 5 days

Statista Daily Data

Infographic: Threads Shoots Past One Million User Mark at Lightning Speed

This chart shows the time it took for selected online services to reach one million users.

Model & API Provider Analysis

artificialanalysis.ai

Model & API Provider Analysis | Artificial Analysis

Comparison and analysis of AI models and API hosting providers. Independent benchmarks across key performance metrics including quality, price, output speed & latency.

LLM Training Process

1

Pre-training

Exposure to extensive textual corpora.

2

Pattern Recognition

Learning language structures and patterns.

3

Fine-tuning

Optimization for specific tasks with targeted data.

Medium

LLM Explained: The LLM Training Landscape

Since the introduction of Transformer model in 2017, large language models (LLMs) have evolved significantly. ChatGPT saw 1.6B visits in…

Dr Alan D. Thompson – Life Architect

Inside language models (from GPT to Olympus)

Understanding Tokens

Definition

Basic units of language processed by LLMs.

Composition

Pieces of words, including spaces and sub-words.

Importance

Token limits influence LLM-generated content depth.

Medium

LLM fine-tuning step: Tokenizing

Tokenization is a crucial phase in fine-tuning the LLM, requiring us to:

NVIDIA Technical Blog

How to Get Better Outputs from Your Large Language Model | NVIDIA Technical Blog

Large language models (LLMs) have generated excitement worldwide due to their ability to understand and process human language at a scale that is unprecedented. It has transformed the way that we…

help.openai.com

What are tokens and how to count them? | OpenAI Help Center

Tokens can be thought of as pieces of words. Before the API processes the request, the input is broken down into tokens. These tokens are not cut up exactly where the words start or end - tokens can include trailing spaces and even sub-words. Here are some helpful rules of thumb for understanding tokens in terms of lengths:

LLM Predictive Nature

Token Prediction

LLMs predict next tokens based on previous ones.

Coherent Responses

Generates relevant and contextual content.

LLM Parameters

Temperature

Controls model creativity and randomness.

Top-k

Limits token selection to top k probabilities.

Top-p

Controls cumulative probability for token selection.

NVIDIA Technical Blog

How to Get Better Outputs from Your Large Language Model | NVIDIA Technical Blog

Large language models (LLMs) have generated excitement worldwide due to their ability to understand and process human language at a scale that is unprecedented. It has transformed the way that we…


LLM Limitations

1

Lack of Understanding

LLMs don't truly comprehend language, facts, or ethics.

2

Outdated Information

Reliance on historical datasets can lead to inaccuracies.

3

Bias Propagation

Models may perpetuate biases present in training data.

4

Ongoing Development

Continuous efforts to address limitations, like real-time internet querying.

Ethical Considerations

1

Authenticity

Concerns about the originality of AI-generated content.

2

Ownership

Questions regarding intellectual property rights of AI creations.

3

Privacy

Data protection and user privacy in AI systems.

4

Responsible Use

Navigating the ethical implications of AI in education.


MIT Sloan Teaching & Learning Technologies

When AI Gets It Wrong: Addressing AI Hallucinations and Bias - MIT Sloan Teaching & Learning Technologies

Understand how AI processes data and generates results and learn to navigate the AI landscape with critical insight to avoid AI's imperfections.

Google

DataGemma: Using real-world data to address AI hallucinations

Introducing DataGemma, the first open models designed to connect LLMs with extensive real-world data drawn from Google's Data Commons.


AI Avator / Voice clone

Relevance to Educators

Tool Accessibility

Students have access to AI tools for assessments.

Detection Challenges

No effective AI-generated content detector currently available.

Widespread Use

AI tools becoming as common as calculators in education.

 


Ithaka S+R

Generative AI Product Tracker - Ithaka S+R

The Generative AI Product Tracker lists generative AI products that are either marketed specifically towards postsecondary faculty or students or appear Stay up-to-date with the emerging tech for higher education. Explore our Product Tracker tool for information, pricing models, key features, and more on the generative AI tools for teaching, learning, and research.


AIGC工具导航

AIGC工具导航 | 生成式人工智能GAI - AI应用工具导航平台！

AIGC工具导航站是一个全网最全,最专注生成式AI人工智能的AI工具导航，包括AI写作、AI绘画、AI设计、AI客服、AI办公、AI营销、AI搜索、AI智能体、AI语音生成、AI视频生成、数字人等生成式人工智能工具。提供一站式AI工具导航服务，帮助用户快速提升工作效率和创作能力。欢迎AI工具创作者提交AI工具网址到AIGC工具导航，共同打造最优质的导航服务，让AI的力量超越想象!

AI in Education: Opportunities

Personalized Learning

AI adapts content to individual student needs.

Automated Grading

Efficient assessment of assignments and exams.

Virtual Tutors

24/7 AI-powered assistance for students.

Data-Driven Insights

Analytics to improve teaching methods and curricula.

AI in Education

Home | AI in Education

AI Advancements Impact

Computer Vision

Enhanced image and video processing capabilities.

Speech Recognition

Improved accuracy in voice-to-text applications.

Multimodal AI

Integration of multiple data types for enhanced interaction.

AI agentic workflows

The use of AI agents to autonomously manage and optimize processes and tasks

Med-PaLM: A Medical Large Language Model - Google Research

Med-PaLM: A Medical Large Language Model - Google Research

Discover Med-PaLM, a large language model designed for medical purposes. See how we developed our AI system to accurately answer medical questions.


Resources

Online Learning

Digital platforms for AI education

Professional Networks

Communities for sharing AI insights

Expert Collaboration

Partnerships with AI specialists

AI Literacy: A Framework to Understand, Evaluate, and Use Emerging Technology

Digital Promise

Homepage – Digital Promise

We work at the intersection of education leaders, researchers, and technology developers to improve learning opportunities...

springshare

LibGuides: Using AI in Education and Research: Home

springshare

LibGuides: Emerging AI Tools for Literature Review: Overview

This guide consolidates the teaching materials for library workshop Emerging AI Tools for Literature Review.

  


Discussion

Potential Applications

Challenges & Opportunities

Responsible AI Use​

Collaboration

Dialogue