The Easy Way to Build Interactive AI Toys for Your Kids

I’m often the first in line to rant about AI, but there is no denying that it can be useful. And though I think parents should be careful about how their kids use AI, the possibilities for interactive entertainment are immense. If you want to get the benefit of that while maintaining control over the AI in question, you can take advantage of Open Toys to build AI toys for your kids.

Open Toys is an open-source system for creating custom toys with LLM (large language model) brains. But crucially, it lets you use whatever offline LLM you want. You aren’t stuck with ChatGPT or Gemini and don’t have to trust that those will interact with your child in an appropriate way. It is important to note that any LLM can say something you won’t like, but at least you’ll get to choose the LLM you want and it won’t update without your knowledge.

I called this a “system,” because it consists of a multiple hardware and software pieces put together. The hardware includes an ESP32-S3 development board that is the “brain” inside the toy and a computer with Apple Silicon to run the backend, including the LLM itself. The backend is built on Tauri and React, which are cross-platform, but it has only been tested on Mx-based Apple computers.

Just stick the ESP32 MDB into a teddy bear or toy of your own making, add a microphone and speaker, and set up the software. You can even clone a voice, if you want to give the toy some unique character. You can set it to only listen when your child presses a button, which alleviates some privacy concerns. The ESP32 will then pass that along to the backend, get a response, and reply out loud.

There are all kinds of possibilities here for making toys tailored to kids as individuals, which is a pretty neat use for AI.