Kyutai, a French AI firm has created a new AI-backed chatbot dubbed “Moshi” that comes packed with features similar to ChatPGT’s now-delayed ‘Advanced Voice Mode’ GPT-4o.
Moshi chatbot is capable of understanding the user’s tone of voice and interpreting it. It can also be used offline. Based on a 7B parameter large language model (LLM) and called Helium, the chatbot is currently available for all and can speak in several accents and 70 different emotional and speaking styles. Moshi can also handle two audio streams at the same time, which means it can listen and talk at the same time.
The AI chatbot, named after the Japanese way of greeting people, has a response time of just 200 milliseconds, amking it swift compared to GPT-4o’s Advanced Voice Mode, which typically takes anywhere between 232 to 320 milliseconds.
Also Read: Netflix's Latest Game Will Take You Back To Your Childhood
Is Kyutai’s Moshi Better Than ChatGPT?
Kyutai says that it focuses on teaching Moshi several nuances and tones of human conversations. To upgrade the voice quality, the company even collaborated with a professional voice artist.
However, unlike GPT-4o, Moshi is quite small and created from scratch in six months by a team of just eight researchers. It was reportedly trained on 1,00,000 synthetic dialogues using Text-to-Speech tech.
Kyutai says its aim is to make the chatbot an open-source project and make the model’s code and framework available to all so that users can safely use the chatbot without compromising on their privacy. While Moshi is faster than GPT-4o, the company says it is a research prototype and is a way for them to put forward the bot’s response time and ability to replicate not only sentences but tones and voices too.
The company is also working on an AI-powered audio identification, watermarking and signature tracking system that will soon be integrated with Moshi. And while this may not be the ChatGPT competitor we have been waiting for, it is a huge leap in the development of open-sourced models that can run offline.
Also Read: Google Wants IPhone Users To Ditch Safari For Its Apps
ChatGPT Mac App Raises Security Concerns For Users
Days after making its debut on the macOS, the ChatGPT app has seen a rather concerning security issue. As per a report from The Verge, the ChatGPT app was storing chats with the chatbot in plain text on the user’s computer.
Pedro Vieto, a developer took to Threads to post a video showing all the chats with ChatGPT were being stored in plain text. “I was curious about why [OpenAI] opted out of using the app sandbox protections and ended up checking where they stored the app data,” he told The Verge.
Anything that is stored in plain text on the computer can be accessed with ease by a hacker or even a malicious app. All the conversations users might have had with ChatGPT were easily readable.