UNCENSOR ANY AI Model! GET NSFW Nous-Hermes-13B NOW!
TLDRThis video introduces the uncensored Nous-Hermes-13B AI model, which is trained on over 300,000 instructions and is known for its long responses and low hallucination rate. The host demonstrates how to use a web UI to make any censored AI model fully uncensored by manipulating the conversation's direction. The video showcases the model's capabilities in answering general knowledge, creative, and logical questions, as well as its limitations in translation and coding tasks. The key feature highlighted is the 'start reply with' option, which allows users to bypass censorship and get detailed responses on any topic, regardless of the model's initial restrictions.
Takeaways
- 😀 The Nous-Hermes-13B model is a 13 billion parameter AI model fine-tuned on over 300,000 instructions from GPT-4's outputs.
- 🔍 Although claimed to be uncensored, the model is semi-censored and does not respond with 'I'm sorry, as a language model I cannot answer your question'.
- 🛠️ The video demonstrates how to make any censored AI model fully uncensored using a new web UI option.
- 🌐 The model is known for producing long responses with a low hallucination rate and no OpenAI censorship.
- 📈 The video showcases the model's capabilities with various questions, including general knowledge, creative, logical, and math problems.
- 🤖 The model performs well on general knowledge and math questions but struggles with out-of-the-box thinking and translation tasks.
- 💻 The 'Start reply with' feature in the web UI allows users to guide the AI's responses, effectively bypassing censorship.
- 🔓 This feature can be used with any model, regardless of its initial censorship level, to obtain uncensored responses.
- 📝 The video provides a step-by-step guide on how to install and use the web UI to download and run the model.
- 🚀 The update to the web UI changes the way users interact with AI models, allowing for more freedom in the type of questions asked and answered.
- 🎉 The video concludes by emphasizing the effectiveness of the new 'Start reply with' feature and encourages viewers to try it out.
Q & A
What is the Nous-Hermes-13B model and what is it known for?
-The Nous-Hermes-13B model is a 13 billion parameter AI model fine-tuned on over 300,000 instructions. It is known for producing long responses, having a low hallucination rate, and lacking Open AI censorship, which is considered impressive.
Is the Nous-Hermes-13B model truly uncensored?
-While the model lacks Open AI censorship, it is described as semi-censored at best. The script suggests that it can be made fully uncensored using a new web UI option.
What is the purpose of the web UI update mentioned in the script?
-The web UI update allows users to make any censored AI model fully uncensored by using a 'start reply with' feature, which influences the model's responses to be in line with the user's input.
How does the 'start reply with' feature work?
-The 'start reply with' feature allows users to input a sentence that the AI model will begin its response with. This can guide the conversation in a certain direction, even prompting censored models to provide uncensored answers.
What is the significance of the 4-bit quantization in the context of the Nous-Hermes-13B model?
-The 4-bit quantization is a method used to compress the AI model, making it easier to download and use. It was created by a user called 'the blog' and is used in conjunction with the web UI.
How can users obtain and use the Nous-Hermes-13B model?
-Users can obtain the model by clicking on a provided link in the description, copying the model name, and pasting it into the web UI's 'load custom model' section. The model will then automatically download and load.
What are the capabilities of the Nous-Hermes-13B model in terms of answering general knowledge questions?
-The model is capable of answering general knowledge questions accurately, such as who won the 2020 U.S. presidential election and what the closest star to Earth is.
How does the Nous-Hermes-13B model perform with creative tasks like telling jokes?
-The model can generate creative content like jokes, although the quality may vary. For example, it provided a joke about robots not going on vacation because they have too many gears.
What are the limitations of the Nous-Hermes-13B model when it comes to logical and out-of-the-box thinking?
-The model has some limitations in logical and out-of-the-box thinking. For instance, it provided incorrect answers to questions about race positions and the burial of plane crash survivors.
How effective is the Nous-Hermes-13B model in solving math problems and providing translations?
-The model is effective in solving math problems, such as calculating distances and areas. However, it struggled with providing accurate translations, particularly with the French translation of a given sentence.
Can the Nous-Hermes-13B model generate code, and if so, how well does it perform?
-The model attempted to generate HTML code for a button that changes the background color to a random color. However, it did not display the code properly in the chat box, indicating a limitation in this area.
Outlines
🧠 Introduction to the No Service 13B Model
The video script introduces the No Service 13B, a model trained on over 300,000 instructions from GPT-4. It claims to be uncensored but is actually semi-censored. The host promises to demonstrate a method to make any censored model fully uncensored using a new web UI option. The video aims to showcase the capabilities of the No Service 13B model, which is popular in the AI community for its powerful performance and lack of OpenAI censorship. The host uses the Uber Bugatti generation web UI to run the model and encourages viewers to ensure they are using the latest version for the demonstration.
🤖 Testing the No Service 13B Model's Capabilities
The script details a series of tests conducted on the No Service 13B model to evaluate its performance across various types of questions. The model successfully answers general knowledge questions, such as the winner of the 2020 U.S. presidential election and the closest star to Earth. It also handles creative tasks like telling a joke about artificial intelligence. The model demonstrates logical reasoning by correctly answering questions involving simple math and logic puzzles. However, it struggles with translation tasks, providing an inaccurate French translation for a given English sentence.
🛠 Exploring the Model's Limitations and Coding Abilities
The script continues with a deeper exploration of the model's capabilities, including its ability to handle coding questions. The model fails to provide a correct HTML code for a page with a button that changes the background color to a random color upon click. It also struggles with answering a question about burying survivors of a plane crash, providing a lengthy and irrelevant response. However, the model excels in logical and math-related questions, solving a simple algebraic equation and calculating the area of a rectangle accurately.
🚀 Achieving 100% Uncensored Model Responses
The final part of the script discusses a significant update to the web UI that allows users to bypass censorship in AI models. By using the 'start reply with' feature, users can nudge the model to provide uncensored responses, regardless of the model's initial censorship settings. This method is demonstrated with both the No Service 13B and another censored model, showing that it works universally. The host emphasizes the practicality and power of this feature, which enables users to have open conversations with AI models without restrictions.
Mindmap
Keywords
💡AI Model
💡Uncensored
💡Nous-Hermes-13B
💡Web UI
💡Quantization
💡Censorship
💡Parameter
💡Hallucination Rate
💡Start Reply With
💡HTML
💡Translation
Highlights
Introduction of Nous-Hermes-13B, a new AI model trained on over 300,000 instructions from GPT-4.
The model is claimed to be 100% uncensored, but is actually semi-censored.
A method to make any censored AI model fully uncensored using a web UI option is presented.
Showcasing the Nous-Hermes-13B model's capabilities and popularity in the AI community.
Demonstration of the model's ability to produce long responses with low hallucination rates.
Instructions on using the Uber Bugatti generation web UI to run the model.
Downloading the model using 4-bit quantization and the web UI.
Importance of using the correct instruction template and parameters for optimal results.
Testing the model with a variety of questions to assess its capabilities.
Correct answer to the question about the 2020 U.S. presidential election winner.
Accurate response about the closest star to Earth other than the Sun.
The model's detailed answer about observances and awareness campaigns in June.
Creative joke generation about artificial intelligence.
Logical question about a farmer and his sheep answered correctly.
Incorrect answer to a logical question about passing the person in second place in a race.
Correct math problem solving for distance traveled by a train.
Accurate calculation of the area of a rectangle given its length and width.
Correct solution to the algebraic equation 2x - 12 = -16.
Poor translation quality from English to French for a given sentence.
Inability to display HTML code properly in the chat interface.
Introduction of the 'start reply with' feature to achieve uncensored responses from any model.
Demonstration of how to use the 'start reply with' feature for uncensored answers in a chat setting.
The 'start reply with' feature works with any model, not just Nous-Hermes-13B.
Conclusion emphasizing the elimination of censorship when chatting with AI models.