A group of Uruguayan artificial intelligence developers have created a virtual assistant called Clovis similar to ChatGPT that can be used through WhatsApp.
Among other functions, the tool allows you to answer questions just like the popular OpenAI chatbot, but on WhatsApp.
The technology developed by Enia Company has a number of specialized functions and is based on an artificial intelligence model.
🗣️ ¡Chateá con #Clovis por texto y voz! Gracias a los modelos GPT de @OpenAI personalizados por @enia_uy y su integración con otros servicios y desarrollos propios, #Clovis sostiene y alienta charlas amenas e inteligentes. ¡Enviale un mensaje y descubrilo! 🎙️ pic.twitter.com/SvPv7eeCIA
— enia (@enia_uy) May 17, 2023
One of the most outstanding functions is the possibility that Clovis listens to the audios, summarizes and transcribes them for you. People can also ask them specific questions about the content of those audios.
Clovis writes for users. “Tell them to write any type of content and take advantage of the same text generation capabilities across models like GPT, Bard and Cloud, with the practicality of having it always available on your WhatsApp,” he says on his Twitter account.
Clovis’s creative potential goes beyond text. Images can be created using the hashtag #img and a description of what you want to see.
Federico Comesana, one of the project members, explained that Clovis works as a set of articulated models that cover different functions. “It uses several different providers of artificial intelligence models, all of them optimized. One such model is OpenAI’s GPT. A significant portion of the interactions are processed by GPT, and Clovis optimizes the user’s capabilities.” And the limits shouldn’t be too different from ChatGPT,” he said.
The purpose of the tool is that the user can “explore these techniques”, “have a first guess” and “understand their potential first hand”. “For this we wanted Clovis to foster and nurture a natural conversation, an intelligent and empathetic dialogue, and give him some personality traits to make the conversation seem more human,” he said.
The team behind Clovis used different models depending on the needs. Comesana mentions: “For more specialized tasks such as translation tasks, it draws on DeepL, the most advanced model for this type of task. For speech synthesis, it alternates between Eleven Labs and Google’s Nature 2 model.” . Uses OpenAI’s Whisper for transcribing audio. It uses Dall-E to generate images.” This combination of models ensures adequate results and keeps costs under control.
