Five AI Experiments Caribbean Parents Can Run With Their Kids This Summer
The Caribbean school summer break has started. Every child in the region will spend the next six to eight weeks doing something. Five free, browser-based AI experiments, built by Google and Code.org, can turn a few hours of that time into the kind of AI literacy the region's schools are still getting ready to teach.
Caribbean parents can teach their children the basics of AI this summer with five free, browser-based experiments: Quick, Draw!, Teachable Machine, Chrome Music Lab, AI for Oceans, and a supervised story-writing session with a chatbot. Each takes under an hour. No coding required. Every experiment builds a specific AI literacy skill Caribbean schools will be teaching by 2027.
The summer window Caribbean parents actually have
Caribbean school summer holidays run from early July to late August or early September, depending on the country. That is roughly six to eight weeks of unstructured time in Jamaica, Trinidad and Tobago, Barbados, Guyana, the Bahamas, Saint Lucia, Grenada, Dominica, Antigua and Barbuda, Saint Vincent and the Grenadines, Belize, Suriname, Haiti, the Dominican Republic, Curaçao, and Aruba. Most of it will be spent on a phone, a tablet, or a television.
The Caribbean's schools are already changing how they think about AI. The Caribbean Examinations Council published a Responsible Generative AI Policy Framework for the regional secondary education system last year. Jamaica launched its UNESCO AI Readiness Assessment on 1 April 2026. The Caribbean AI Association is publishing a CARICOM AI policy framework within 90 days. By the September 2027 school year, most Caribbean children will be in classrooms where AI is a standing topic, not a novelty.
Parents do not need to wait for the school system to catch up. The five experiments below are the fastest, cheapest, safest way to give a Caribbean child, ages 4 to 15, a real feel for how AI actually works. Every one of them was built for education. None cost money. None require software installation. All of them run in a browser on any laptop, phone, or tablet with an internet connection.
Do the experiments with the child, not for the child. AI literacy is built by doing, asking questions, and getting things wrong. Every experiment below includes discussion prompts. The prompts matter as much as the tool. A child who has drawn ten sketches and then answered "why did the AI get confused?" understands more about AI than most adults.
Five experiments to try before school restarts
The five below are ranked by starting age, from youngest to oldest. Most families can complete all five in a single week, spending 20 to 60 minutes on each. Doing them in order builds a natural learning curve: recognise, train, create, evaluate, collaborate.
Google's Quick, Draw! shows a child how AI recognises patterns
The child gets a prompt like "cricket bat", "mango", or "hummingbird". They have twenty seconds to draw it. As they draw, an AI, trained on millions of human doodles, guesses out loud what it thinks they are drawing. It is fast, funny, and unmistakably an AI. Children as young as four laugh at the wrong guesses. Children ten and up start noticing patterns in what the AI gets right and what it does not.
Open quickdraw.withgoogle.com in any browser. No sign-up. No download. Play three or four rounds. After each round, look at the drawings from other players around the world that trained the AI. Ask the child what they notice.
Notice which prompts fit Caribbean life easily (mango, palm tree, guitar, boat) and which are harder to relate to (skyscraper, snowman, ferris wheel). Ask the child what happens to an AI that was trained mostly on drawings from other parts of the world. That is a real conversation about training data.
How did the AI know what you were drawing so fast? What did you draw first that helped it guess? Why did it get confused on that one? What would happen if only children in Kingston, Bridgetown, or Port of Spain had drawn the training pictures?
Google's Teachable Machine lets a child train a real AI classifier
This is the single best five-minute introduction to machine learning available to any Caribbean family. The child points the device camera at an object, records twenty or thirty examples, then does the same for a second object. They press "train". Within seconds, they have a working AI that recognises the difference. Testers running the tool with 23 children aged 6 to 15 reported an average of seven minutes from opening the URL to a working model, with every single child succeeding within ten minutes.
Open teachablemachine.withgoogle.com. Choose "Image Project", then "Standard image model". Point the phone or laptop camera at object A (say, a mango). Hold the record button for 20 samples. Do the same for object B (say, an ackee, or a school shoe). Press "Train Model". Test it live by holding each object up to the camera.
Train it on things that matter to the child. Callaloo versus lettuce. Cricket bat versus tennis racket. Two family members with different uniforms. Local fruit against imported fruit. Then deliberately try to fool it. What does the AI struggle with? What does that teach you about its training data?
What did the AI need from you to learn? What happened when you showed it something it had never seen? What would a fair training set look like for our house, our school, our community? If a company trained a fruit-recognition AI only on apples and oranges, what would it get wrong in the Caribbean?
Chrome Music Lab teaches how AI meets creativity
Chrome Music Lab is a collection of fourteen music experiments built by Google Creative Lab. Song Maker, the most popular, is a colourful grid where a child taps blocks to compose a melody and a rhythm. Kandinsky turns any drawing into music. Melody Maker plays back what you compose. It works on any tablet, does not require reading, and gives a four-year-old their first musical composition within a minute of opening the tool.
Open musiclab.chromeexperiments.com. Start with Song Maker for younger children, Kandinsky for artistic children, or Melody Maker for children who already play an instrument. Let them compose freely for the first fifteen minutes, then set a challenge (compose a song for a birthday, for a hurricane warning, for a football match).
Try to build a soca, reggae, calypso, zouk, kompa, or bouyon rhythm on the Song Maker grid. Notice which patterns are easy and which are harder. Then let the child export the song and share it with a musical relative or teacher. Ask a musician in the family what they hear. This starts a conversation about what music AI can and cannot capture about Caribbean rhythm.
Can a computer make music? What is missing when a computer makes music? What is the difference between a song the AI helped you compose and a song you sing yourself? If AI wrote your favourite song tomorrow, would you still love it? Why or why not?
Code.org's AI for Oceans teaches training data, bias, and ethics
Every Caribbean child understands the ocean. Code.org's AI for Oceans uses that connection to teach one of the most important concepts in machine learning: bias in training data. The child trains an AI to sort fish from trash so a virtual robot can clean the sea. As they train, they discover that the AI learns exactly what they teach it, including the mistakes. This is the same problem that produces biased hiring algorithms, biased medical diagnostics, and biased facial recognition in the real world. Over one million students have completed the activity globally.
Open studio.code.org/s/oceans. Follow the five levels in order. The first level trains the AI to sort fish from trash. Later levels introduce non-fish sea creatures, then invite the child to define their own categories. Do not skip the videos between levels; they are short and clearly explain what is happening inside the AI.
The Caribbean sits at the front line of marine plastic pollution. After the activity, discuss how a similar AI could help clean beaches from Negril to Nassau, from Grand Anse to Grande Riviere, from Grace Bay to Bathsheba. Ask the child which local coastline they would send the AI robot to first, and what training data it would need to be useful there.
What happens if the person training the AI has never seen a Caribbean beach? What is the AI missing when it decides what counts as trash? Who should choose the training data for an AI that will be used across many countries? What did the AI get right and what did it get wrong when you tested it?
Co-write a Caribbean short story with Claude, Gemini, or ChatGPT
This last experiment introduces the child to how large language models work, how they get things wrong, and how to prompt them well. The parent opens a mainstream chatbot on their own account and sits alongside the child. The child provides the story idea. The AI provides a first draft. The child edits it. The pair go back and forth until the story feels right. Along the way, the child learns three of the most important skills in AI literacy: prompting, fact-checking, and creative direction.
Use the parent's account, not a child's account. Most mainstream chatbots require users to be 13 or older, so stay signed in as the parent and treat this as a joint activity. Start with a clear prompt: "Write me the opening paragraph of a story about a child in Kingston who finds an old radio that plays songs from the future." Read the AI's response together. Ask the child what they would change. Rewrite the prompt. Try again.
Ask the AI for a story set in your parish, town, or island. See what details it gets right (Blue Mountain coffee, Kalinago heritage, Carnival, Junkanoo, Divali) and which it gets wrong or invents. Fact-check three claims from the story with the child. Discuss the difference between an AI that sounds confident and an AI that is correct.
How did the AI decide what a Caribbean setting looked like? Did it get anything wrong about our country? How did we rewrite the prompt to make the story better? Should we trust an AI that sometimes makes up facts? How would you check if what an AI told you was true?
Build a seven-week summer schedule with the family
The five experiments do not need to happen in a single day. A seven-week rhythm, matching the length of a typical Caribbean school break, gives each experiment time to breathe. Week one is Quick, Draw!, run as a family game night. Week two is Teachable Machine, run as a weekend afternoon project. Weeks three and four spread Chrome Music Lab across several sessions with musical family members. Week five is AI for Oceans, ideally paired with a real beach clean-up. Weeks six and seven are the storytelling collaboration, ending with a finished short story to keep, print, or read to grandparents.
| Week | Experiment | Time | Family Setting |
|---|---|---|---|
| Week 1 | Quick, Draw! | 30 min | Family game night; take turns drawing while others guess along with the AI. Easy |
| Week 2 | Teachable Machine | 45 min | Saturday afternoon; train a classifier on household items, then run "fool the AI" challenges. Easy |
| Week 3 | Chrome Music Lab (Song Maker) | 60 min | After school-holiday camp; each child composes a "song of the summer" to share at family dinner. Easy |
| Week 4 | Chrome Music Lab (Kandinsky & Melody Maker) | 45 min | Quiet Sunday activity; older children compose, younger children draw pictures that become music. Easy |
| Week 5 | AI for Oceans | 60 min | Rainy day inside; if the weather is good, follow the activity with an actual beach clean-up. Medium |
| Week 6 | AI storytelling: draft the story | 45 min | Parent and child at the kitchen table, one laptop, one chatbot open on the parent's account. Medium |
| Week 7 | AI storytelling: fact-check and finish | 45 min | Read the draft aloud, fact-check three claims, edit together, print or save the final version. Medium |
Rules for AI and kids that actually work
Every Caribbean parent should follow four rules when introducing children to AI, whether the tool is Quick, Draw! or a chatbot. These are not restrictions. They are the difference between a summer of AI literacy and a summer of AI passivity.
Sit with the child. AI is one of the very few technologies where the child learns more with a parent in the room, asking questions, than they would on their own. The parent does not need to know the answers. Curiosity is enough. Say what you notice. Ask what they notice. Let them explain the tool back to you.
Use the parent's account for chatbots. Most mainstream chatbots require users to be 13 or older, and even then, some content is not appropriate for children. When using ChatGPT, Claude, Gemini, or any similar tool, the parent stays signed in and treats the session as a joint activity. This is the same principle as a driving lesson: the child steers, the adult owns the vehicle.
Always ask what the AI got wrong. Every AI in every experiment above will make mistakes. The mistakes are the lesson. A child who has watched Quick, Draw! misread their bicycle as a stethoscope has learned more about how AI actually works than a child who watched a video about neural networks.
Connect the AI to Caribbean life. The most valuable moments in every experiment are when the tool encounters something recognisably Caribbean and either succeeds or fails. That is where the child feels the difference between an AI trained on Silicon Valley data and an AI trained on the world the child actually lives in. That difference is the seed of a future Caribbean AI industry.
Never let a child sign up for an AI service on their own. Never share the child's name, school, address, or photo in a chatbot conversation. If a chatbot produces content that is scary, sexual, violent, or otherwise inappropriate, close the tab, take a screenshot, and report it to the provider. Google's AI Experiments collect minimal data and are ad-free; other providers vary. Check the settings for chat history retention on any mainstream chatbot before use.
Why this summer matters more than the last one
AI use in Caribbean classrooms will be normal by the September 2027 school year. The Caribbean Examinations Council's Responsible Generative AI Policy Framework has already gone out to CXC-participating secondary schools across Antigua and Barbuda, Barbados, Belize, Dominica, Grenada, Guyana, Jamaica, Saint Lucia, Saint Vincent and the Grenadines, and Trinidad and Tobago. The CARICOM-UNDP Joint Regional AI Programme (2026 to 2030) is funding public sector deployments in Jamaica, Trinidad and Tobago, and beyond. The children who arrive at that school year already having trained a classifier, composed a song, and fact-checked a chatbot will be ahead of ninety per cent of their classmates.
The children who arrive having only watched AI-generated content on a phone will not be behind because they were less capable. They will be behind because their summer was passive when it could have been active. Six to eight weeks is a long time. Five experiments is a short list. Every Caribbean parent, in every one of the sixteen countries this newsletter covers, can start this weekend.
The Caribbean's next generation will not fall behind on AI because the region lacks talent. If they fall behind, it will be because their summers were passive when they could have been active, and because their parents waited for the school system to catch up before starting the conversation at home. Five experiments. Six weeks. Every parent, from Nassau to Paramaribo, from Willemstad to Port au Prince, can start this weekend. Caribbean AI Newsletter — The Classroom Shift, 29 June 2026
About Caribbean AI
Caribbean AI is the official directory of artificial intelligence companies, labs, and innovators in the Caribbean. We exist to connect startups, enterprises, and researchers driving the region's AI growth.
We track who is building AI in the Caribbean, what is being built, and what the region needs to compete. From Jamaica to Curaçao, from Guyana to the Bahamas, we document the people, organisations, and ideas shaping the Caribbean's AI future.