My 8-year-old daughter is obsessed with how Alexa works and she's been asking me how the tablet knows its her face when she unlocks it. Shes a sharp kid but her school is really lagging behind on anything tech-related so I've been trying to find something we can do at home over the weekends. We're based in a small town outside Philly and the local library doesn't have much beyond basic how to use a mouse classes which is pretty frustrating.
I did some digging and found a few things but nothing feels quite right:
I've got about $25 a month to spend if it's a subscription model but I really want something interactive where she can maybe train a tiny model or see how logic paths work in a visual way. Is there anything out there that's actually designed for elementary ages that isn't just a glorified YouTube video? I want her to actually grasp the concepts without getting bogged down in the heavy math stuff she hasn't even seen in school yet...
To add to the point above: Google Teachable Machine is a decent, free option for visualizing training sets. Its very direct. For actual project building, Makeblock mBlock 5 Software works better than standard Scratch for younger kids. It has specific blocks for face and speech recognition that feel pretty professional. Both tools help explain her tablet questions without needing any complex math.
I've seen kids get frustrated when software is too abstract, so you might want to consider something hardware-based that bridges the gap. I would suggest looking into STEMpedia Quarky Ultimate Kit for AI and Robotics. It uses a block-based interface which is less intimidating than standard Scratch extensions. You should probably make sure to check if her tablet supports their PictoBlox app before buying tho, as older hardware sometimes struggles with the camera processing required for the face recognition modules. One tip: be careful with tools that rely entirely on cloud processing. They can lag depending on your internet speed. A few things to verify:
Wait really?? Thats actually super helpful. I always thought it was the other way around.
Just caught this thread and honestly, the biggest hurdle I've found with kids this age is moving past the black box stage where everything just feels like magic. In my experience, showing them that AI is basically just pattern matching with specific datasets is far more effective than any abstract code lesson. Over the years, testing several vision sensors has shown that the DFRobot HuskyLens Gravity AI Vision Sensor is usually the best entry point for specs-minded kids. It uses a specialized Kendryte K210 chip to perform face recognition and object tracking locally on the hardware. Connecting it to a micro:bit lets her see the confidence percentages and bounding boxes in real time. This helps explain those logic paths she's curious about without needing the heavy calculus she hasn't learned yet. Quick tip: when she starts training it, have her change the background or lighting. It's a perfect way to show how bias enters a model when data isn't diverse. Way more interactive than just clicking a screen tho...
Helpful thread 👍