ReddRadar
What is SmartestChild?
This AI chatbot, named SmartestChild, utilizes a Large Language Model and runs completely within the user's browser via WebGPU. It mimics a 2000s-era AIM chatbot, responding with lowercase text, emojis, and internet abbreviations. Importantly, no user data is transmitted to any server.
All conversations are locally cached, maintaining user privacy. The system is powered by Apache TVM and MLC Relax Runtime, with the Vicuna model trained by LMSys. Note that WebGPU is only supported in Desktop Google Chrome 113 and the initial model download may take several minutes, although it will be cached afterwards.
Features
- Local Execution: Runs entirely within the browser using WebGPU.
- Data Privacy: No data is sent to the server.
- Conversation Caching: Conversations are stored in local storage.
- 2000s Aesthetic: Responds with lowercase, emojis, and 2000s internet abbreviations.
- Powered by Apache TVM: Uses Apache TVM and MLC Relax Runtime.
Use Cases
- Casual conversations
- Experimenting with a locally run Large Language Model
- Testing WebGPU capabilities
