Nous Research, a noted AI research group, has announced the launch of its first user-facing chatbot, Nous Chat. The chatbot provides users with access to its Hermes 3-70B large language model (LLM), which is a variant of Meta’s Llama 3.1. This introduction marks a significant development for Nous Research, which has previously released several open source models and efficient AI training techniques. However, users who wanted to use these models were required to download and run code on their machines, a process that could be time-consuming and costly.
The new platform utilises a standard interface similar to other popular AI chatbots like ChatGPT. It features a text entry box at the bottom for user input and a large space above for outputs. Nous’s announcement on social media platform X highlighted that since the release of their first Hermes model over a year ago, there was a significant demand for a platform where people could experience these models directly.
The design of Nous Chat is described as reminiscent of early PC terminals, featuring vintage fonts and a user-friendly toggle between dark and light modes. Nous Chat also includes suggested prompts such as “Knowledge & Analysis” and “Creative Writing,” aimed at enhancing user interaction.
Initial assessments of Nous Chat have been positive, particularly noting the speed of response, with answers typically generated in seconds. The chatbot also has the capability to produce links to web sources for citations, though instances of “hallucination,” where the chatbot invents non-existent sources, have been noted. Despite its innovative approach, Nous Chat has implemented certain content restrictions, an approach that echoes developments by other AI model providers like OpenAI.
The Hermes 3-70B model that powers Nous Chat contains a knowledge cutoff of April 2023, which limits its ability to provide information on current events. This aspect presents a competitive disadvantage compared to other AI models actively engaged in providing up-to-date information.
Nevertheless, the platform has quickly attracted the attention of AI jailbreakers such as Pliny the Prompter. This indicates challenges Nous Research faces in maintaining content restrictions while meeting their aim of offering unrestricted AI models.
Although Nous Chat presently lacks several advanced features, such as image analysis and interactive code display, the launch is viewed as an intriguing experiment. The group plans to introduce additional features to enhance its functionality as a potential alternative to current market leaders in the realm of AI chatbots. The future development and feature expansion of Nous Chat could potentially fulfil Nous Research's objective of providing less restricted AI models for users.
Source: Noah Wire Services