HugginChat: An efficient, free, open "ChatGPT" that doesn't want your personal data

HugginChat: An efficient, free, open “ChatGPT” that doesn’t want your personal data

Since the end of 2023, the ChatGPT fever has taken over the internet. And unless you’ve been living in low Earth orbit for the past few months, you’ve probably been tempted to use artificial intelligence in the form of a chatbot or conversational agent, whether these be OpenAI’s ChatGPT, Anthropic’s Claude, Copilot from Microsoft or more recently Google Gemini (formerly Bard). It must be said that these algorithms are really efficient, sometimes so much so that they replace traditional search engines with smoother and more “human” interaction.

However, the conversations with these chatbots are far from trivial and there is nothing very clear about the use of the data you send during your exchanges to the companies that offer the services, often for free. They likely feed the massive databases needed for training AI and their evolution. However, the global enthusiasm around AI offers the opportunity to try many solutions, and we recommend that you test the conversation agent available through the Hugginface platform.

Different models available

Created by the French, Hugginface lets you test numerous AI models right in your browser, whether processing audio data, performing image recognition, or generating images or music. The platform’s conversation agent is called HugginChat.

To use the Hugginface chatbot, you do not need to create an account.

To test the chatbot, nothing could be easier: just go to the URL You can then choose to use the AI ​​in guest mode or create an account, which will retain your conversation history and certain preferences and settings. As for preferences elsewhere, note that it is possible to use different LLM models (great language model) and therefore obtain different responses depending on the selected neural networks. If the model of the French company Mistral is selected by default (mistralai/Mixtral-8x7B-Instruct-v0.1), you can switch to the LLM developed by Meta (meta-llama/Llama-2-70b-chat -hf) or OpenChat. These models have the advantage that they have been developed in open source and are therefore freely accessible to developers.

Depending on your questions and the purpose of your use, the different models will likely yield different answers. It is up to you to select the one that suits your needs.

Stay in control of your data

Still in the preferences section you have the option to refuse sharing conversations with the authors of the models. Please note that this is the case by default, so you will need to intervene in the Application settings before starting a conversation. However, keep in mind that if you share your conversations with the AI, your conversations may be used to improve the models.

Like all AI models, the models used by HugginChat were trained with time-limited data. The answers to your current questions may therefore turn out to be outdated, as you can see with our question about the Prime Minister of France.

But HugginChat can query the internet to supplement its information and from that moment on the agent becomes much more relevant.

Friendly assistants

Like the ChatGPTs available in the OpenAI store, HugginChat also offers the use of personalized assistants provided by the community. You can chat with an agent specialized in code generation, image generation and why not… an Italian grandmother!

More interestingly, you can create your own custom wizard. For this it is necessary to resort to some fast which allows you to more precisely define the scope of action of your “entity” to get more accurate and relevant answers.

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *