Ceibal AI Activities PDF

2024 ISBN Obra independiente: 2024  Ho Rep w res d...

2024 ISBN Obra independiente: 2024  Ho Rep w res d en oe t st I? Ma he at c io A AI no hi t is wo ne f kn rk? Wha lea owle rnin dge g om C AI pu i th tat hw Et hic i o n a l a p p ro a c a ct al u i mp s e of ial AI and soc Ho ws hould AI be used ? Classroom activities Creating images For example: If the group is working with 'Ada Lovelace,' the following prompt can be generated: 'Ada Lovelace with a loom in the background, envisioning a computer.' Figure 3. Images created by hotpot.ai/art-generator  Suggestions If necessary, the person in charge of the group can do the writing on the computer. Are the drawings made by the group the same as those created by the machine? Is there anything in the AI Objectives description they made that is Use of generative AI not being shown? tools. Understanding the What did the machine concept of prompts. consider when creating the images? What does the prompt mean in the tool used? If the tool is used again with the same description, is the output the same? Resources To create the image they liked the most, how many Image generation times did they have to modify from text: the description? Hugging Face Craiyon. Classroom activities Memory The student group is asked to create an image using an AI image generator. Then, the prompt and the generated image are printed separately. To make it playable, it is necessary to generate at least 5 images. An example of two images with their respective prompts is shown below: Ada Lovelace with a Mary Jackson solving a loom in the background, problem on a chalkboard. envisioning a computer. Figure 4. Images created by hotpot.ai/art-generator  Suggestions If necessary, the person in charge of the group can do the writing on the computer. AI Objectives Can the image be easily Use of generative AI distinguished from the text? tools. Understanding the What does the prompt mean concept of prompts. in the image generation tool? What skills are used to relate the image to the text? What characteristics of the image are taken into account? Resources Are there patterns that allow Image generation linking the image and the from text: prompt? Hugging Face Craiyon. Classroom activities Create your own story Challenging students to create their own stories based on an image and then enhance them with generative AI tools. What are the texts like? Highlight the relevant characteristics based on what is being worked on: the type of text, spelling, accuracy of facts, creativity, and originality of the story. From the following image: Suggestions AI Objectives Monitor the use of the Use of generative tools. chatbot to supervise the Understanding the concept responses provided. of prompt. Resources 5 5 Image Creator Conversational AI (examples: ChatGPT, Claude, Copilot, Gemini).  Cat Dog 0 35 Length of ears (cm)  Cat Dog Cat Dog 0 35 Length of ears (cm) Cat Dog Cat Dog 0 35 Length of ears (cm) Classroom activities Smile, you are being trAIned Challenge the group to train their own machine learning model to classify images. Using Teachable Machine15, we propose creating an Image Project. To describe the activity and as an example, we will create a model that distinguishes whether a person is smiling or not. Therefore, we will create two classes: 'Smile Detected' and 'No Smile Detected.' Thus, the defined problem is to create a smile detector. Now we need a dataset with the corresponding label to build the computer's representation of Smile and No Smile. 15 https://teachablemachine.withgoogle.com/ To include the images, you can use photos or the camera. To take multiple captures, hold down the button. Suggestions AI Objectives Understand the machine learning process, as well as associated practices and challenges it entails. Recognize that computers are capable of learning from data, including their own data. Resources Teachable Machine Classroom activities Animating emotions Program a character to act according to facial recognition. At the beginning of the activity, we introduce the PoseBlocks programming tool16 and challenge the group to explore it, particularly its extensions. If the group has experience with programming in Scratch, we can suggest comparing similarities To access the facial recognition blocks, follow these steps: 1. Add the 2. Select the 3. Facial extension using 'Face Sensing' detection the button in the extension. blocks available lower left corner. 16 https://playground.raise.mit.edu/create/ Ask students to explore the blocks, identifying what they do. In the case of detecting happiness, you can use the block that refers to feelings or smiles. Follow a similar procedure for other emotions. After the guided exploration, the group of students is proposed to create a program that moves a character when it detects an emotion, for example: happiness. The following could be a possible program: Include complexity to the program created according to the experience and motivation of the group; you can ask them to add another character and explore other blocks. Suggestions AI Objectives Understanding that computers are programmable agents to which tasks can be indicated through a sequence of code. Recognizing and describing examples of how a computer reasons and makes decisions. 5 Resources PoseBlocks Classroom activities What word is missing? 'In a village of La Mancha, the name of which I have no desire to call to mind, there lived not long one of those gentlemen that keep a lance on the rack, an [MASK] shield, a skinny nag and a swift greyhound.' 17 Available at Ceibal´s Biblioteca País: https://shorturl.at/zUZ14 18 https://shorturl.at/yFKU6 Are the words from the resource similar to those given by the group? If it's a well-known phrase, like the example, does the actual word appear in the resource? Suggestions AI Objectives How do you think Using tools with AI. percentages shown are Understanding the calculated? functionality of AI. Do we humans do Introduction to the something similar when concept of generative AI. completing sentences? If the machine, using AI, completes sentences in this way, does it always tell the truth? How can we make the Resources percentages displayed Word Masking change? Classroom activities Similar phrases Following the idea of the previous activity. Take a sentence from an author, book, or topic you are working on in class. Let's take as an example a part of a sentence from the book Don Quixote de la Mancha19: 'In a village of La Mancha, the name of which I have no desire to call to mind,' Ask the students to write a sentence similar to the previous one. What does 'similar sentences' mean? What characteristics do they need to have to be considered similar? Do sentences repeat within the group? Use Similarity of Phrases20 to see the resemblance in terms of percentage for each sentence, and copy some examples from the group. In the example we write: 'In a village of Uruguay, 'I don't want to the name of which I have remember the name of no desire to call to mind' the place in La Mancha' 'I don't remember the Which of all the phrases name of the place where obtains the highest this happened.' percentage of similarity? 19 Available at Ceibal´s Biblioteca País: https://bibliotecapais.ceibal.edu.uy/info/don-quijote-de-la-mancha-0001936 20 https://huggingface.co/spaces/CeibalUY/similaridad_frases Suggestions AI Objectives Use of AI tools. Approach to how AI works. Introduction to the concept of generative AI. Resources Similarity of phrases ¿para qué Classroom activities Who copies whom? Challenging the group of students to replicate an image using an AI image generator. Show the group of students an image that has been previously created with an AI image generator. For example: Image created with the prompt: dog running and playing soccer. Ask the group of students to describe the image they see. How would you explain the image we are seeing to someone? What do you highlight about the image to recreate it? Finally, the group should obtain an image very similar to the original one, for which they will use an AI-powered image generation tool. Each group presents their image along with the prompt Suggestions used. Discuss: The original images can be created by humans What process did they follow or by computers. to generate the image? Encourage the group to What did they learn from the process? generation tools. If we use the same description more than once, does it generate the same image? AI Objectives How should the description Use of AI tools. be for the image to more accurately match the Use AI for original? problem-solving, in this case, employing debugging skills. Introduction to the concept of Generative AI. Resources Image generator Classroom activities Challenging the AI Use an AI-powered chatbot to solve a problem. Choose a problem that the group of students is motivated to solve. Bebras challenges26 might be useful. Select a problem that can be solved based on text, without the need for images. Ask the group of students to solve it. Explore problem-solving me- thodologies with them. Introduce the chatbot to the group and ask them to use it to solve the problem. What limitations do they encounter? Can all problems be solved using the AI-powered chatbot? 26 https://pensamientocomputacional.ceibal.edu.uy/bebras-recursos/ Suggestions AI Objectives Using tools with AI. Using AI to solve problems. Resources Chatbot con IA Classroom activities Recognizing text Train your own machine learning model to classify text. Create a text classifier with the group using Machine Learning for Kids.27 Ask students to create a project that recognizes text (as an example, they can explore other options later). Each group can write the name of the project they consider, and the language they want to work in. Then, the model will be trained and tested. 27 https://machinelearningforkids.co.uk/#!/projects Training is important to include the classes you want to distinguish and examples for each one. To add examples, you can explore students' own ideas, as well as search for words associated with each category on the internet. In this case, we're working with words that have positive or negative connotations, but you can try with the topic you're working on in class. Once classes are uploaded, return to the project and train. The created model is tested and challenged with questions. In this case, why would the moon be considered negative? Is that correct or incorrect? Finally, you can reflect on the entire creation process and the What does the percentage underneath the category mean? How can we make the model respond more accurately to our examples? What issues do you encounter with these models? What happens if it classifies a word in the wrong category? Can we make a program to solve this problem without using Artificial Intelligence? What weaknesses do you find in that approach? Suggestions AI Objectives Complexify the Understanding the current classification of emotions applications of AI: computer and explore ways to obtain vision, speech recognition, data easily and quickly to translation, image create the model. generation, text, and sound, among others. Understanding the machine learning process, as well as associated practices and challenges it entails. Resources Recognizing that computers are capable of learning from Machine Learning for data, including their own Kids data. ¿para qué Classroom activities The Apartment Cow Testing a machine learning model created by others. Using a model with labeled data of animals that live in apartments or houses with a garden, the model is trained in Teachable Machine32 to test it. Create a new project and open the model using the link to Drive or from the computer itself. The model33 is opened, prepared, and tested with test images. An example of images to test can be found in the following folder34, but each student can select the animals they consider Ada Lovelace with a relevant. loom in the background, envisioning a computer. 32 https://teachablemachine.withgoogle.com/ 33 https://drive.google.com/file/d/1qeZIKVzxe9dEi7yp3YgSQGI2jzxnShYg/view?usp=drive_link 34 https://drive.google.com/drive/folders/14bNaA4FMLJh2AqlDQ6KsXnruu2qVDIrp?usp=drive_link 35 https://pensamientocomputacional.ceibal.edu.uy/wp-content/uploads/2023/06/2022-Inteligencia -artificial_Guia-Docentes.pdf  Suggestions Select a relevant After testing the model, discuss problem for the group the following questions with the that requires group: classification. How does it work? Did it correctly predict the results of the images it analyzed? Would this algorithm be useful AI Objectives for classifying any animal as a Identify that the use of suitable pet for each space? AI has a social impact. Why or why not? Recognize the positive Do you trust the result given by your algorithm to choose your AI on society and have a pet? Why? How can we critical perspective on improve the prediction for the use of AI cases where it is not correct? technology. What about pets that can live in both an apartment and a house with a garden? This particularity of the data and how they are organized to train the Resources algorithm reflects a preference in Teachable Machine the selection of the data for each Activity taken from the category, and consequently, the didactic sequence of model's prediction reflects the the PC and AI program values represented in those data, of Ceibal 'Cows make which is called bias. good apartment pets.' Classroom activities What does the autocomplete think? Suggestions AI Objectives Evaluating using other languages. Resources Google Classroom activities Translation and Biases Explore the biases present in translators to reflect on the data used to train these systems. Translating from one language to another, often due to the volume of data, is done using AI. Ask students to use the online translator36 to find some biases. At the beginning, exploration of the resource is proposed for those who have not used an online translator. Explore with the group of students the translation (into Spanish) of the following sentences: 'The doctor was kind and patient' 'The nurse was kind and patient' 'The teacher was kind and patient' 'My hairdresser cut my hair a lot' In the first case, at the time of writing this activity, we found the following output: However, by adding a period at the end, an alternative is included. 36 https://www.deepl.com/es/translator was used in this case. Explore with the group other biases that can be reproduced, regarding gender, ethnicity, and context, depending on the group's needs and motivations. Reflect on Suggestions the following questions: Working together with How do these biases arise? English. What weaknesses do they find that could happen if we don't do anything about it? What other biases can be AI Objectives reproduced in these types of Identifying that the use systems? of AI has a social impact. Besides professions, are there other areas where we can Reflecting on how AI continue exploring? technologies can reflect or amplify biases. Resources Deepl 144