New GitHub Project 'WeClone' Builds Digital Twins from Chats
A new open-source project on GitHub, dubbed WeClone, is gaining traction among developers. According to LΣҒΔ𝕽ΩLL 🇮🇱, this tool aims to create a ‘digital twin’ from chat conversations. The project has already garnered significant attention, boasting around 16.8k stars and 1.4k forks.
WeClone is presented as an end-to-end solution for building a personalized AI chatbot. It facilitates exporting chat histories, cleaning sensitive data like phone numbers, emails, credit cards, and IP addresses using Microsoft Presidio, and then fine-tuning a language model locally. This process allows the resulting bot to mimic the user’s communication style.
The project supports image inputs and utilizes the Qwen2.5-VL-7B-Instruct model with LoRA by default. LΣҒΔ𝕽ΩLL 🇮🇱 notes that the developers acknowledge it’s still a work in progress, with performance not yet finalized. For a more convincing digital twin, users might need more data and significant GPU resources.
What This Means For You
- If you're concerned about the potential misuse of AI for impersonation or the security of your personal data, be aware of tools like WeClone. Review your chat history privacy settings and consider what information you share online, as it could potentially be used to train such models.