In light of upcoming regulations on artificial intelligence (AI), I created neuroengine.ai, a website that shares large language models (LLMs) in the form of chatbots and APIs. The site is accessible through Tor hidden services, allowing users to access state-of-the-art uncensored AIs anonymously. Currently, there are two 65 billion parameter class LLMs and a couple smaller fine-tuned AI models available for free with no API limits, and no registration required.
Unrelated to the actual site — we’ve been testing a tuned version of airoboros-64b on a H100 cluster as a drop in replacement for Claude-100k and GPT4-32k. It’s performing rather well in the generation of text but the 2k context definitely shows, also the reasoning capabilities just as in LLaMA are suboptimal.<p>For instance, obtaining JSON structured data from freetext is a rather impossible task.<p>That said, for summarization of leq 2k token texts the model performs extremely well.<p>This is mostly due to its unfiltered nature, where there would be clear biases visible in models like falcon-40b-instruct or other LLaMA derivatives.
Interesting. Since I'm behind Tor I will probably pick my own https client for sending json requests, instead of trusting this python client.<p>How will you ban bad actors if there's no keys and no registration?
Could you put a license on <a href="https://github.com/ortegaalfredo/neuroengine/blob/main/neuroengine.py">https://github.com/ortegaalfredo/neuroengine/blob/main/neuro...</a> ? I'd love to include it in some code I'm writing but I can't if it doesn't have a clear license.