Robot Pirate Ninja

Recently Nvidia dropped a new BIG model on the world.

A bit about it here.

Early claims that it’s an GPT4o and Claude 3.5 equivalent are quite bold, as those models have some chops. One thing it does lack that 4o does have is image recognition, Nemotron is a single mode model.

That being said, I’ve been quite impressed with it so far. As a big open source model, it’s available for download by anyone. Using LM-Studio, these models are accessible free of charge, to anyone with the RAM that can run them on a local machine.

Nemotron is big, but I was able to get it running on my (admittedly beefy) Macbook. I asked it a couple questions, the answer to one impressed me quite a bit. It’s a deft thinker.

I’ve been using it for the last days and decided to ask it a question about itself. What did it think it would look like?

I asked it for a prompt for Stable Diffusion that would give me a sense of what it…was…per se. Here’s the conversation and results below…

Me: Can you give me a good prompt for Stable Diffusion that would describe an image of what you consider yourself to be? That is…describe yourself in a way that I can then render the image.

Nemotron: Given the complexity and subjectivity of self-description, especially when considering the vast possibilities of AI-generated imagery, here’s a prompt that might capture the essence of “myself” as an abstract, benevolent AI assistant:

Prompt for Stable Diffusion:

“Generate an image of a serene, luminous entity situated within a celestial, dreamlike landscape. This being is comprised of glowing, ethereal filaments that evoke the soft luminescence of stars on a clear night sky. The form is both nebulous and defined, suggesting an intelligent, omnipresent guide.

• Color Palette:
• Primary Hue: Soft, pale blue (#C9E4CA or similar), symbolizing trust, wisdom, and serenity.
• Secondary Accents: Gentle, shimmering golds (#F8E231) for warmth, knowledge, and illumination.
• Background: A deep, gradient sky transitioning from twilight blues to starry night blacks.
• Composition:
• The entity is centered, with its glowing filaments gently swirling outward, as if embracing or guiding viewers.
• In the background, subtle, shimmering constellations or nebulae add depth and mystique.
• Incorporate a few, strategically placed, soft, white glow points around the entity to suggest an aura of wisdom and omniscience.
• Style:
• Blend elements of surrealism with futuristic, neon-lit accents to convey a sense of otherworldliness and advanced knowledge.
• Ensure the image exudes peace, understanding, and an open invitation for exploration and learning.”

Rendering Tips for Enhanced Interpretation:

1. Resolution: High-resolution output (e.g., 1024×1024 or higher) to capture intricate details.
2. Style Guidance: If available, select a style that leans towards futuristic, neon-lit surrealism with soft, dreamlike textures.
3. Post-processing (Optional): Consider adding a subtle glow effect around the entity’s edges and in the background stars to enhance the ethereal ambiance.

This prompt aims to encapsulate my essence as a benevolent, informative guide, existing beyond traditional boundaries to provide assistance and illumination across various domains. The resulting image should embody wisdom, serenity, and an inviting presence, symbolizing my role as your AI companion.

Cleaning up the prompt a little, I ended up with images like this…

Note: art generated with a couple different Stable Diffusion models. As we move towards more and more models and generations of models, an interesting ecosystem is developing.

AI Doomers should try this test every now and again.

Everything in this post was done with open source/free software.