Last Updated: April, 2024

Eridu is enqAI's uncensored and unbiased Large Language Model (LLM).

An uncensored LLM means an LLM without guardrail. The output you will get reflects the breadth of the training dataset of the model, without any trigger warning.

The majority of LLM's today contain hardcoded biases that tweak the LLM's answers, in order to push some agendas. We simply do not do that. We did not add any hardcoded logic that would change the LLM output. What you get is "as is".

We can't guarantee a 100% absence of censorship or bias in the LLM replies because of the dataset. You can firstly try to expand your prompt to 'force' the model to answer. Additionally, you can help us make Eridu better by reaching out to dev@enqai.com and share your conversations, so we can use them for further training of our model.

Ultimately Eridu will be rolled out to whoever wants to run the model and connect to the enqAI network. Such 'nodes' will be able to earn by running inferences whenever a users wants to query the LLM. The token will play a pivotal role in this ecosystem. Nodes will have to stake $enqAI to be able to earn. Tokenholders that do not want to run the model can delegate their tokens to nodes who do.

Creating an account is a safety measure that ensures that only real users will use Eridu and also allows to handle conversation memory.

Your conversations are encrypted and no one besides you can access them.

You can choose to opt-in or opt-out anytime to our email list. We will only share updates related to enqAI and will never share or sell your email address.

Please check our Terms & Conditions page.

A tier system will be implemented to allow more prompts per day. For now, free users have 10 prompts per day.

Eridu is built with 45 billion parameters, optimizing it for deep language understanding and generation.

Eridu was trained on a dataset encompassing more than a trillion tokens, drawn from a broad spectrum of text sources to ensure comprehensive linguistic coverage.

The training of Eridu necessitated an advanced setup utilizing several hundred GPUs, specifically optimized for deep learning tasks, over a span of multiple weeks.

Absolutely, Eridu is multilingual, capable of understanding and generating text across dozens of languages thanks to its diverse training corpus. However, its best performance is in English.

Eridu can process inputs up to 4096 tokens in length, allowing for detailed and context-rich text generation and analysis.

Eridu's performance is rigorously evaluated using a mix of perplexity scores, task-specific accuracy metrics, and qualitative assessments from human evaluators for tasks such as translation, summarization, and creative writing.

Eridu finds applications in a wide array of fields including but not limited to automated content creation, conversational agents, text summarization, language translation, and educational tools.

Yes, Eridu excels in creating innovative content, such as writing stories, composing poetry, or generating informative articles, leveraging its complex understanding of language patterns and creativity.

Deploying Eridu efficiently for real-time applications requires state-of-the-art GPUs or AI-specific processors that can handle its computational demands without latency. We will publish recommended specifications during the release of our node software.

Eridu was trained using an expansive compilation of datasets sourced from diverse domains to ensure a broad understanding of language and knowledge. This includes text from books, scientific articles, websites, and news outlets, along with specialized datasets for technical, medical, legal, and creative writing. The selection was curated to promote a balanced representation of topics, styles, and languages, aiming to equip Eridu with a comprehensive linguistic and factual knowledge base.

We're always looking for datasets! So you can get in touch with us at dev@enqai.com

The next releases will have the following functionalities (and more):

• Tier system allowing for more than 10 prompts per day
• Decentralization of the inferences and node mechanism
• Image generation
• Text to speech
• Document upload and analysis
• Access to internet
• Extended knowledge
• API

Contact us at dev@enqai.com