Resources for Learning about AI Technologies

Please note: This is not intended to be an exhaustive list of articles or articles that reflect the government’s preferences.

Example and Thought Process for an Elementary Classroom
Example and Thought Process for Middle and High School Teams
Creating Unique Slideshows with AI
Learning About Advanced/New AI Tools
Advancing Experienced Python Coders' AI Skills
Online Video Resources
Website Programs for Learning to Code

Example and Thought Process for an Elementary Classroom
AI tools have age restrictions (found in the end user license agreement or other requirements for tool use), so the educator will need to implement the tool use based on input from the students.

  • Design an AI tool that would help students find lost items.
    • User Interface for Reporting/Querying
      • Natural Language Input: Students can verbally describe or type details about their lost item.
      • Example: "I lost my blue water bottle, the one with the stickers, probably near the gym after lunch."
    • Structured Data Collection: The AI prompts for key details:
      • Item Type (e.g., backpack, phone, jacket, water bottle)
      • Distinguishing Features (e.g., color, brand, unique markings, stickers)
      • Last Known Location (e.g., classroom, library, gym, specific area)
      • Time Lost (e.g., "this morning," "yesterday afternoon")
      • Student's Contact Information (for notification purposes)
    • Centralized Lost & Found Database:
      • "Lost" Item Log: 
        • Stores all reported lost items with their descriptions and student contact info.
      • "Found" Item Log
        • School staff (or potentially other students) can log items they find, including location found and a detailed description.
        • Images of found items could also be uploaded.
    • Matching Algorithm
      • Uses keyword matching, fuzzy logic, and potentially image recognition (if images are provided for found items) to cross-reference "lost" reports with "found" items.
      • Prioritizes matches based on item type, unique features, and spatial/temporal proximity.
    • Proactive Matching Alerts 
      • If a match is found, the AI immediately notifies the student.
      • Example: "Good news, [Student's Name]! I think I found your blue water bottle! A bottle matching your description was found near the gym entrance this afternoon. Please check with Mrs. Chen in the main office to retrieve it."
    • Search Guidance
      • If no immediate match is found, the AI offers helpful advice.
      • Example: "I haven't found a match for your [item] yet, [Student's Name]. Keep an eye out in your usual spots or try retracing your steps. I'll keep looking and let you know if anything turns up!"
    • Follow-up
      • The AI could periodically check in with the student about the status of their lost item until it's found or marked as unrecoverable.
    • Underlying Technology and Integration:
      • Natural Language Processing (NLP): For understanding student queries and generating friendly responses.
      • Machine Learning (ML): To refine safety risk assessment, improve matching accuracy, and personalize interactions.
      • Cloud Infrastructure: For database hosting, processing power, and scalability.

Example and Thought Process for Middle and High School Teams
Distinguishing Crops from Weeds and Analyzing Related Issues
This scenario involves a progression of AI applications, moving from classification to analysis and then to potential mitigation.

Step 1: Identifying Crops or Weeds

  • Concept: This requires Classification using Machine Learning.
  • Potential Tools/Approaches:
    • Python with Libraries: This is a robust approach. Students and teachers can use libraries like:
      • Scikit-learn: For traditional machine learning models.
      • TensorFlow/Keras or PyTorch: For more advanced deep learning models like LSTMs (Long Short-Term Memory networks) which are helpful for image classification tasks.
      • Pandas: For data manipulation and cleaning.
      • NumPy: For numerical operations.
      • Matplotlib: For creating graphs and visuals from your dataset.
    • No-Code/Low-Code Platforms:
      • Roboflow
      • Microsoft Azure Machine Learning Studio (designer): Offers a drag-and-drop interface to build machine learning pipelines, including classification.

Step 2: Analyzing Issues Arising from Classification
Concept: This involves AI-powered data analysis and natural language processing (NLP) to interpret the implications of the classifications.

  • Potential Tools/Approaches:
    • Large Language Models (LLMs) like Gemini or ChatGPT:
      • Input: A set of labeled images of crops and weeds from step 1 and information about your location like the USDA plant hardiness zone, soil conditions, weather conditions, or other useful data about your location.
      • Prompt: Ask the LLM to analyze potential issues. For example: "What are implications of these weeds for my local crop or farming operations?", "What economic consequences could arise?", "How might these weeds affect my neighbors?" LLMs can synthesize information and identify interconnected problems based on their vast training data.
    • Data Visualization Tools (though not strictly AI, crucial for analysis):
    • Roboflow, FiftyOne, Tableau, Matplotlib, PowerBI: Visualize identified plants in various images, potentially distribution of crops and weeds over time, and other statistics around your dataset. This can help in formulating better questions for the AI analysis tools.

Step 3: Discovering Ways for AI to Fix Issues

  • Concept: This is where AI moves from analysis to prescriptive recommendations and potentially autonomous action. This is often the most complex and research-intensive part.
  • Potential Tools/Approaches (more conceptual and research-oriented):
    • LLMs for Solution Brainstorming:
      • Prompt: "Given the identified weeds in the crops, what AI-driven solutions could be proposed to address these problems?"
      • LLMs can suggest solutions like: recommending specific herbicides that will likely be successful given your site conditions and weed concentrations, precision agriculture techniques like smart herbicide application systems that only spray weeds and not crops, smart irrigation systems (AI-controlled water delivery) that waters crops but not weeds, or something entirely novel and not described here.
    • Reinforcement Learning (RL) (Advanced): For complex, dynamic problems where an AI agent needs to learn to make decisions to optimize an outcome, RL could be considered. For instance, an RL agent could be trained to determine the optimal timing and location for deploying targeted herbicide applications, learning over time from weather patterns, crop growth stages, and previous successes or failures. This approach enables AI systems to go beyond static rules and adapt to changing conditions in the field, making precision agriculture more responsive and efficient.

Optimization Algorithms (often integrated with AI): AI methods can be used to find optimal solutions to complex problems. For example, AI-based tools could help optimize irrigation schedules based on sensor data about soil moisture and crop needs, or determine the most efficient routes and timing for drone-based weed surveillance. These techniques can be particularly useful in resource-constrained environments where maximizing yield with minimal input is essential. By integrating optimization algorithms into their AI pipeline, learners can explore how algorithmic thinking directly supports viability, productivity, and cost-effectiveness in modern agriculture.

Creating Unique Slideshows with AI
For those new to AI and looking to create dynamic presentations, the key is to use tools that simplify the content generation and design process.

What to do:

  • Gather Content: First, ensure all photos and information are downloaded and organized. Have a clear idea of the presentation's topic and key messages.
  • Choose an AI Presentation Tool: These tools are designed to take your raw content and turn it into visually appealing slides.

Sample Tools
While this is not intended to be an exhaustive list or one that reflects the government’s preferences, some examples of tools that may be used for this competition are:

  • Gamma: This tool allows users to input text or even just a topic, and it generates a presentation outline and slides. You can then easily add your downloaded photos and refine the content.
  • Canva's Magic Design: Canva has integrated AI features where you can describe the presentation you want to create, and it will suggest designs and even generate some content. You can then upload your own images and text.
  • Beautiful.ai: This tool focuses on design and uses AI to ensure your slides look professional and consistent.

How to get started:

  • Start with a clear prompt: When using these tools, be as specific as possible about your topic, the audience, and the key takeaways.
  • Iterate and refine: AI-generated content is a starting point. Review, edit, and personalize the slides with your unique insights and downloaded materials.

Learning About Advanced/New AI Tools
For those looking to dive deeper into AI and its problem-solving capabilities, focusing on Generative AI and Machine Learning Platforms will be beneficial.

Advanced AI Concepts to Explore:

  • Large Language Models (LLMs): Tools like ChatGPT (OpenAI), Gemini (Google AI), and Claude (Anthropic) can be used for complex tasks like:
    • Ideation and Brainstorming: Generating creative solutions to problems.
    • Content Generation: Drafting reports, summaries, or even code snippets.
    • Data Analysis (Preliminary): Asking questions about datasets (though not a substitute for dedicated data analysis tools).
    • Simulations: Creating hypothetical scenarios to analyze potential outcomes.
  • Diffusion Models (for image/media generation): Tools like Midjourney, DALL-E 3 (OpenAI), and Stable Diffusion are changing creative fields. While primarily for image generation, understanding their underlying principles (how they learn from data to create new outputs) can inspire novel problem-solving approaches, such as:
    • Prototyping visual solutions: Designing interfaces or products.
    • Generating synthetic data: Creating diverse datasets for training other AI models.
  • No-Code/Low-Code AI Platforms: Platforms like Google's Vertex AI (with its AutoML features) and Microsoft's Azure Machine Learning allow users to build and deploy machine learning models without extensive coding. These help with understanding the workflow of building an AI solution from data to deployment.

Applying these tools to solve a problem:

  • Problem Identification: Encourage students and teachers to identify a real-world problem that involves data, analysis, or creative generation. Examples:
    • Optimizing school bus routes.
    • Predicting student performance.
    • Generating unique lesson plan ideas.
    • Designing a public awareness campaign.
  • Proposing Solutions: Once a tool is understood, the focus should be on how its specific capabilities can directly address the identified problem. For instance, how could an LLM help analyze student feedback to identify common themes, or how could a diffusion model generate visual aids for a complex science concept?

Advancing Experienced Python Coders' AI Skills
For experienced Python coders, the path to creating "awesome projects" involves deepening their understanding of core AI concepts, mastering advanced libraries, and focusing on real-world applications.

Guidance for Skill Advancement:

  • Deepen Understanding of Machine Learning Fundamentals:
    • Beyond Scikit-learn: While scikit-learn is useful for traditional ML, encourage a deeper dive into the mathematical and statistical underpinnings of algorithms.
    • Focus on Model Interpretability (Explainable AI - XAI): Understanding why a model makes certain predictions is crucial for trust and debugging. Libraries like LIME, SHAP, and eli5 are helpful for this.
    • Bias and Fairness in AI: Explore how to detect and mitigate bias in datasets and models, a critical aspect of ethical AI development.
  • Master Deep Learning Frameworks:
    • TensorFlow 2.x and Keras: Useful for building and deploying neural networks. Focus on custom layers, callbacks, distributed training, and deployment with TensorFlow Serving.
    • PyTorch: Another widely-used deep learning framework, particularly popular in research. Understanding its dynamic computational graph can be very beneficial.
    • Transformers Library (Hugging Face): For those interested in NLP, mastering the transformers library is key for working with state-of-the-art LLMs, fine-tuning them for specific tasks, and deploying them.
  • Explore Specialized AI Fields:
    • Computer Vision: OpenCV, Pillow, and deep learning models (CNNs) for image recognition, object detection, and segmentation.
    • Natural Language Processing (NLP): Beyond transformers, explore NLTK, spaCy for text preprocessing, entity recognition, and more advanced topics like sentiment analysis, topic modeling, and summarization.
    • Reinforcement Learning (RL): Libraries like OpenAI Gym and Stable Baselines3 provide environments and algorithms for experimenting with RL, which is powerful for sequential decision-making problems.
  • Focus on Deployment and MLOps:
    • Building a model is only half the battle. Learning how to deploy models into production (e.g., via APIs, cloud platforms like AWS SageMaker, Google Cloud AI Platform, Azure ML) is crucial.
    • MLOps (Machine Learning Operations): Understand the pipeline for continuous integration, continuous delivery, and monitoring of AI models. Tools like MLflow, Kubeflow, and Docker/Kubernetes are relevant here.
  • Project Ideas for "Awesome Problems":
    • Personalized Learning Assistant: An AI that adapts learning materials based on a student's progress and learning style.
    • Environmental Monitoring System: Using satellite imagery and ML to detect deforestation, water pollution, or track extreme weather pattern indicators.
    • Smart City Optimization: AI for traffic management, energy consumption optimization, or waste management.
    • Generative Art/Music/Code: Creating novel content using advanced generative models.

How to advance:

  • Open-Source Contributions: Contribute to existing AI projects on GitHub to learn from experienced developers and improve coding practices.
  • Kaggle Competitions: Participate in Kaggle competitions to work on real-world datasets and benchmark skills against others.
  • Build a Portfolio: Create end-to-end projects that demonstrate diverse skills, from data collection and model training to deployment and visualization.
  • Read Research Papers: Stay updated with the latest advancements by reading papers from conferences like NeurIPS, ICML, and ACL.
  • Collaborate: Work with others on complex projects to learn different perspectives and problem-solving approaches.

Resources for Elementary Educators: Scratch coding: https://lab.scratch.mit.edu/face/

Online Video Resources

Website Programs for Learning to Code
Please note: This is not intended to be an exhaustive list of articles or articles that reflect the government’s preferences.

https://scratch.mit.edu
Scratch is a free website where youth can learn to code by making fun games and animations.

https://www.codemonkey.com/
CodeMonkey is a game-based online platform where youth can learn to code through interactive challenges and projects.

https://ww.create-learn.us/ai-for-kids
Create & Learn offers online AI courses for youth, teaching AI concepts through interactive projects, coding, and hands-on activities.

https://teachablemachine.withgoogle.com
Teachable Machine is a free web tool by Google that lets anyone create machine learning models without coding, allowing users to train computers to recognize images, sounds, and poses.

https://code.org/
Code.org is a free educational website that aims to teach computer science to students of all ages, offering coding lessons and activities for elementary, middle, and high school levels through its "Hour of Code" program and other resources.

https://www.tynker.com
Tynker is an online platform that teaches coding to youth through interactive courses and games, offering a range of programming languages from block-based to text-based like JavaScript and Python.

https://machinelearningforkids.co.uk/
Machine Learning for Kids is a free educational website that introduces children to artificial intelligence and machine learning concepts through hands-on coding projects and interactive tutorials.