2.3 All About AI

 What is Artificial Intelligence (AI)?

At its core, AI is about creating computer systems that can perform tasks that typically require human intelligence. This includes things like:

• Learning: Acquiring information and rules for using it.
• Reasoning: Using those rules to draw conclusions.
• Problem-solving: Finding solutions to issues.
• Perception: Understanding sensory input (like seeing or hearing).
• Language understanding: Processing and interpreting human language.

How Does AI Work?

AI achieves these abilities through various techniques, but here are some of the most important ones:

• Machine Learning (ML): This is a subset of AI where systems learn from data without being explicitly programmed. Instead of hard-coded rules, they use algorithms to identify patterns in data and make predictions or decisions.
  • Supervised Learning: The algorithm is trained on a labeled dataset, meaning the desired output is provided along with the input data. For example, training an algorithm to recognize cat pictures by showing it many pictures of cats labeled as “cat.”
  • Unsupervised Learning: The algorithm is given unlabeled data and must find patterns or structures on its own. For example, grouping similar customers together based on their purchasing behavior.
  • Reinforcement Learning: The algorithm learns by interacting with an environment and receiving rewards or penalties for its actions. This is often used in robotics and game playing.
• Neural Networks: Inspired by the structure of the human brain, these are complex networks of interconnected “neurons” that process information. Deep learning is a subfield of machine learning that uses very deep neural networks to solve complex problems like image recognition and natural language processing.
• Natural Language Processing (NLP): This enables AI to understand, interpret, and generate human language. It’s crucial for applications like chatbots, machine translation, and sentiment analysis.
• Algorithms: At the heart of AI are algorithms, which are sets of instructions that enable computers to process data, recognize patterns, and make decisions.

Key Components and Processes

1. Data: AI systems need data to learn and improve. The quality and quantity of data are crucial factors in the performance of an AI.
2. Algorithms: These are the sets of rules or instructions that the AI system follows to process the data.
3. Models: Machine learning algorithms create models, which are representations of the patterns they find in the data. These models are then used to make predictions or decisions on new data.
4. Training: The process of feeding data to an ML algorithm to create a model.
5. Inference: The process of using a trained model to make predictions or decisions on new, unseen data.

In essence, AI works by enabling computers to learn from data, identify patterns, and make decisions or predictions with minimal human intervention.

SCENARIO ACTIVITIES

It’s awesome you’re diving into the world of AI and machine learning! To make these concepts more tangible, here are some simple activities and scenarios perfect for students and ICT club members:

1. The “Human Sorting Machine”

• Concept: Introduction to Algorithms and Automation
• Activity: Divide the students into small groups.
  • Provide each group with a mixed set of objects (e.g., a bag of mixed buttons, a set of cards with different shapes, or a list of unsorted numbers).
  • Challenge each group to develop a set of step-by-step instructions (an algorithm) to sort the objects according to specific criteria (e.g., sort the buttons by color, the cards by shape, or the numbers in ascending order).
  • Have each group test their algorithm by having one student act as the “sorting machine” and follow the instructions precisely.
• Learning Outcomes:
  • Understand the concept of an algorithm as a sequence of instructions.
  • Appreciate how automation works by breaking down a task into simple steps.
  • Recognize that different algorithms can achieve the same result with varying efficiency.

2. “If-Then-Else” Challenge

• Concept: Rule-Based Systems
• Activity: Present the students with a series of scenarios with clear conditions and outcomes. For example:
  • “If it is raining, then take an umbrella. Else, wear a hat.”
  • “If the light is green, then cross the street. Else, wait.”
  • “If a student’s score is above 90, then give them an A. Else, check the score again.”
  • Ask the students to write down the “if-then-else” rules for each scenario.
  • You can make it more interactive by having students act out the scenarios, following the rules they’ve defined.
• Learning Outcomes:
  • Grasp the basic logic of decision-making in AI.
  • Learn how to express rules in a structured format.
  • See how simple rules can be combined to create more complex systems.

3. “Guess the Pattern”

• Concept: Pattern Recognition (a fundamental aspect of machine learning)
• Activity:
  • Prepare a sequence of data with a hidden pattern (e.g., a series of numbers, shapes, colors, or sounds).
  • Present the first part of the sequence to the students and challenge them to guess the next element.
  • As students make guesses, provide feedback (correct or incorrect) and reveal more of the sequence if needed.
  • Discuss how students identified the pattern and what clues they used.
• Learning Outcomes:
  • Understand that machines can learn by recognizing patterns in data.
  • Develop pattern recognition skills.
  • See the importance of data in making predictions.

4. “Teachable Machine Challenge”

• Concept: Introduction to Machine Learning
• Activity: Use Google’s Teachable Machine to introduce a practical machine-learning activity.
  • Students can work in groups to train a machine learning model to recognize different categories, such as:
    • Different hand gestures
    • Various types of objects
    • Different musical instruments
  • Students gather data, train the model, and test its accuracy.
• Learning Outcomes:
  • Gain hands-on experience with machine learning.
  • Learn the basic steps involved in training a model.
  • Understand the impact of data quality on the model’s performance.

5. “AI in Everyday Life” Brainstorming

• Concept: Applications of AI
• Activity: Start a class discussion about where students encounter AI in their daily lives.
  • Encourage them to think beyond obvious examples like robots and self-driving cars.
  • Examples:
    • Recommendation systems on streaming services
    • Smart assistants on phones
    • Spam filters in email
    • Facial recognition in security systems
  • Create a mind map or a list of all the examples.
• Learning Outcomes:
  • Become aware of the widespread use of AI in everyday technology.
  • Start thinking about the potential benefits and challenges of AI.
  • Develop critical thinking and discussion skills.

HOW DOES AI WORK?

Earlier, you learned how artificial intelligence is being used in different areas to make an extraordinary impact on our daily lives. Let’s go a little deeper into what it is and how it works.

True artificial intelligence is not quite here yet. There doesn’t yet exist a system that completely thinks and acts like a human. When we think of AI in our everyday lives, we are really thinking about machine learning. 

When we talk about AI in this curriculum, we’ll really be talking about two subsets of Artificial Intelligence, Machine Learning and Generative AI.

What are they?

Machine Learning

Machine Learning is a subset of AI where a machine (computer) “learns” to identify patterns so it can make predictions.

Examples

Machine learning is indeed a subset of AI where machines learn to identify patterns and make predictions. Here are 12 common examples of how it’s used in everyday life:

1. Spam filtering: Email providers use machine learning to identify and filter out spam messages, keeping your inbox clean.
2. Youtube predictions: Youtube can predict the next video you might like to watch
3. Facebook Identification: Facebook can identify your face in an image.
4. Fraud detection: Banks and credit card companies employ machine learning algorithms to detect unusual transactions and prevent fraud.
5. Medical diagnosis: Machine learning aids doctors in diagnosing diseases from medical images (like X-rays or MRIs) and patient data.
6. Online advertising: Targeted ads on websites and social media are powered by machine learning, showing you products and services you might be interested in.
7. Stock market prediction: Traders use machine learning algorithms to analyze market trends and make predictions about stock prices.
8. Weather forecasting: Machine learning helps improve the accuracy of weather forecasts by analyzing vast amounts of meteorological data.
9. Self-driving cars: Autonomous vehicles rely heavily on machine learning to perceive their surroundings, make driving decisions, and navigate roads safely.
10. Speech recognition: Virtual assistants like Siri, Alexa, and Google Assistant use machine learning to understand and respond to your voice commands.
11. Product recommendations: E-commerce websites use machine learning to suggest products you might like based on your browsing history and purchase behavior.
12. Language translation: Online translation tools like Google Translate use machine learning to translate text and speech between different languages.

Generative AI

can generate text, images, and sounds. It uses Large Language Models to be able to create content based on lots and lots of existing data. 

ChatGPT and DALL-E as prominent examples of generative AI. Here are 12 common examples of how this technology is being used:

1. Text Generation (like ChatGPT): Creating articles, blog posts, social media content, and even scripts.
2. Image Generation (like DALL-E and Midjourney): Producing original images from text descriptions, used in art, design, and marketing.
3. Code Generation: Assisting developers by generating code snippets, or even entire programs, in various programming languages.
4. Music Composition: Creating new musical pieces in different styles and genres.
5. Voice Synthesis: Generating realistic-sounding speech for virtual assistants, audiobooks, and dubbing.
6. Video Creation: Producing short videos, animations, or special effects.
7. 3D Model Generation: Designing 3D models for use in games, virtual reality, and product design.
8. Drug Discovery: Creating novel molecular structures for potential new drugs.
9. Materials Science: Designing new materials with specific properties.
10. Fashion Design: Generating new clothing designs and patterns.
11. Game Development: Creating game assets, levels, and storylines.
12. Data Augmentation: Generating synthetic data to train other machine learning models, especially when real data is scarce.

Let’s dive deeper into Machine Learning.

MACHINE LEARNING HAS THREE MAIN PARTS

DATASET

FINDS PATTERNS WITH LEARNING ALGORITHM

PREDICTION!

Source: “Learning about Artificial Intelligence: A hub of MIT resources for K-12 students”, MIT Media Lab

DATASETS

How AI Uses Data

• Datasets: AI algorithms learn from massive amounts of data, called datasets. These datasets can include text, images, audio, video, sensor readings, and more.
• Learning from Data: AI algorithms analyze the data to identify patterns, relationships, and correlations.
• Pattern Recognition: Once the AI identifies these patterns, it can use them to make predictions or decisions on new, unseen data.
• Prediction: When presented with new data, the AI applies the learned patterns to predict an outcome or classify the data into a specific category.

Where Does AI Get Its Data?

You’re right; a significant portion of the data comes from our everyday interactions with technology. Here’s a slightly expanded list:

• Every Google search query
• Words typed into emails, messages, and documents
• Questions and commands given to virtual assistants (Alexa, Siri, Google Assistant)
• Data from connected devices (smart home devices, wearables)
• Taps, swipes, and other interactions on smartphones and tablets
• Online purchase history and browsing behavior
• Social media connections, posts, likes, and shares
• Streaming activity (songs, videos)
• Location data from GPS and mobile devices
• Sensor data from devices (accelerometers, gyroscopes)
• Cookies and tracking data from websites
• Publicly available data (e.g., from government agencies, research institutions)

The YouTube Example

Let’s analyze how YouTube uses AI:

• Dataset:
  • User viewing history (videos watched, watch time)
  • User interactions (likes, dislikes, comments, shares, subscriptions)
  • Search queries
  • Video metadata (title, description, tags)
  • User demographics (age, location – if available)
  • Video popularity and trends
• How YouTube Learns:
  • YouTube’s AI algorithms analyze the dataset to identify patterns in user behavior and video characteristics.
  • For example, it learns which videos are often watched together, which videos users like or dislike, and which videos are similar to each other.
• What YouTube Predicts:
  • Which videos a user is likely to watch next.
  • Which videos to recommend on the homepage, in search results, and in the “Up Next” sidebar.
  • Which videos are trending and popular.

Activity: AI in Action

It’s great you’re encouraging exploration of those websites! Here’s a breakdown of how they relate to AI and the data they use:

• Instrument Playground:
  • AI Concept: This likely uses a combination of computer vision (to analyze the image) and audio synthesis (to generate the sound).
  • Data Needed: A dataset of images paired with corresponding audio recordings of instruments or soundscapes. The AI learns to associate visual elements with specific sounds.
• AutoDraw:
  • AI Concept: This uses machine learning, specifically image recognition and classification.
  • Data Needed: A large dataset of user doodles paired with labeled, professionally drawn images. The AI learns to recognize the features of a doodle and match them to the closest known drawing.
• X Degrees of Separation:
  • AI Concept: This likely uses machine learning for image analysis and similarity comparison.
  • Data Needed: A dataset of a large number of artworks with metadata describing their visual features (e.g., color palettes, shapes, styles). The AI learns to identify similarities between artworks based on these features.

REFLECTION

Now that you have a glimpse of what it takes to create artificial intelligence, you might want to think about the problem you are solving, and how AI might be useful.

Can you think about your problem and possible solution in terms of the three parts of AI – data, pattern, prediction?

How would you address all three parts in your solution?

Can you think about your problem and possible solution in terms of the three parts of AI – data, pattern, prediction?

How would you address all three parts in your solution?

Can you think about your problem and possible solution in terms of the three parts of AI – data, pattern, prediction?

How would you address all three parts in your solution?

 Here’s a framework for approaching a problem with AI, focusing on data, patterns, and prediction:

1. Define the Problem

Clearly state the problem you want to solve. What is the specific issue you’re trying to address? What are the inputs and desired outputs?

2. Break Down the Problem into AI Components

• Data:
  • What data is needed to solve this problem?
  • Where can you get this data?
  • What type of data is it (text, images, numbers, etc.)?
  • How much data is required?
  • How will you collect, store, and preprocess the data?
• Pattern:
  • What patterns or relationships exist within the data that can help solve the problem?
  • What type of AI/machine learning model is suitable for finding these patterns (e.g., classification, regression, clustering)?
  • How will you train the model to recognize these patterns?
• Prediction:
  • What kind of prediction or output do you want the AI to generate?
  • How accurate does the prediction need to be?
  • How will the predictions be used?
  • How will you evaluate the performance of the predictions?

3. Develop a Solution

• Based on the data, pattern recognition approach, and desired prediction, design an AI-powered solution.
• This might involve selecting a specific machine learning algorithm, building a neural network, or using a combination of AI techniques.

4. Implementation and Evaluation

• Implement your solution, train your AI model, and test its performance.
• Evaluate the results: How accurate are the predictions? How well does the solution solve the problem?
• Iterate and refine your solution based on the evaluation results.

Example: Problem – Traffic Congestion

Let’s apply this framework to the problem of reducing traffic congestion in a city:

• Problem: Reduce traffic congestion during peak hours.
• Data:
  • Real-time traffic flow data from sensors and cameras
  • Historical traffic data
  • Road network information (maps, lane configurations)
  • Weather conditions
  • Public transport schedules
  • Incident reports (accidents, construction)
• Pattern:
  • Identify recurring patterns in traffic flow (e.g., rush hour bottlenecks, congestion hotspots)
  • Correlations between traffic flow and factors like time of day, weather, and events
• Prediction:
  • Predict traffic flow in different areas at different times
  • Predict the impact of potential interventions (e.g., changing traffic light timings)
• Solution:
  • Develop an AI system that uses the data to predict traffic conditions and optimize traffic light timings in real-time.
  • The system could also provide drivers with alternative route suggestions and inform them about potential delays.

REVIEW OF KEY TERMS

• Artificial Intelligence (AI): The broad field of creating machines or computer programs that can perform tasks that typically require human intelligence. This includes abilities like learning, problem-solving, and decision-making.
• Machine Learning (ML): A subfield of AI that focuses on enabling machines to learn from data without being explicitly programmed. Instead of being given specific rules, machines learn to recognize patterns in data, allowing them to make predictions or take actions.
• Generative AI: A type of AI that can create new content, such as text, images, music, and videos. It learns the underlying structure and patterns of existing data and then uses that knowledge to generate something original.
• Large Language Model (LLM): A powerful type of generative AI model that is trained on massive amounts of text data. LLMs can understand, generate, and manipulate human language, making them capable of tasks like writing articles, translating languages, and engaging in conversations.
• Datasets: Large collections of structured data that are used to train AI models. The quality and size of the dataset significantly impact the performance of the AI.

ADDITIONAL RESOURCES

Want to explore some more cool AI?

• Labs.Google: This website showcases Google’s latest AI experiments and projects. You can find interactive demos and tools that highlight cutting-edge AI research in areas like natural language processing, computer vision, and machine learning. It’s a great place to see AI in action and get a glimpse into the future of the technology.

• NVIDIA AI Playground: NVIDIA is a leading company in AI hardware, and their AI Playground provides a platform to experiment with various AI models and tools. You can often find demos related to image generation, style transfer, and other visual AI applications.