Hi, good morning
This is to share with you that our research group made publicly available the beta version of *the chatbot for the Portuguese language based on open LLMs, Evaristo.ai https://evaristo.ai/ * One of such LLMs is Gervásio 8B https://huggingface.co/PORTULAN/gervasio-8b-portuguese-ptpt-decoder, we developed for Portuguese and we are releasing also now, and which is the LLM that you find active by default when arriving at this chatbot.
Though you may not be proficient in Portuguese, on-the-fly translators will likely help you to get its content and its basic functioning. *We invite you, and your colleagues and students, to visit it and try out this first test version. We welcome any help you can give us in testing it, any feedback and any suggestions you can share with us.*
It has a unique set of features, among others: it is an open AI chatbot for the Portuguese language; it is multi-model and multi-heteronym, as well as being agentic, multi-tool and multi-modal; it does not track its users or pass on their content to third parties, safeguarding user privacy and ownership of their content.
You'll find the presentation of its motivation in this press release https://evaristo.ai/assets/pressRelease_EvaristoAI.pdf(in English), that can be complemented by the more complete description in the section about https://evaristo.ai/about (in Portuguese)
The current open LLMs available are typically between 10 and 100 times smaller than the top-of-the-range closed LLMs used in commercial chatbots, so the costs associated with training and operating them are much lower. The performance of open LLMs, however, is much more satisfactory than this linear disproportion would suggest. They therefore have an excellent ratio of performance quality versus cost, and are a viable option for fully autonomous generative AI services focused on concrete use cases.
In this context, we see this chatbot as a milestone in the democratization of generative technology for the Portuguese language, through open LLMs, by encouraging more and more organizations to move forward with their own AI services, rooted in their own computers, and focused on their concrete use cases.
Have a nice day,
António