🤖 AI/ML - Neural Texturing Challenges
Neural texturing faces a fundamental problem: low per-neuron computation density. Traditional ReLU neurons perform a dot product, one activation, and output a single scalar, making them inefficient for complex functions. Purpose-built kernels can often compute the same structure more efficiently.
Key Points:
Neuron Computation Density: Traditional ReLU neurons have low per-neuron computation density, making them inefficient for complex functions.
Kernel Efficiency: Purpose-built kernels can often compute the same structure more efficiently than traditional ReLU neurons.
MLP Scaling: MLP solutions don't scale well with the number of channels stored in each latent texture and the number of desired decoded output RGB channels.
🔗 Resources:
- Original post ↗
- Original source ↗
- Neural Texturing ↗
- Brief description of neural texturing challenges
🤖 AI/ML - NNTC Weight Scaling
NNTC's total number of weights scales based on the number of channels stored in each of the two latent textures and the number of desired decoded output RGB channels. MLP solutions don't scale well this way, making their decoders "lobotomized".
Key Points:
Weight Scaling: NNTC's total number of weights scales based on the number of channels stored in each latent texture and the number of desired decoded output RGB channels.
MLP Scaling Limitations: MLP solutions don't scale well with the number of channels and desired decoded output RGB channels.
Decoder Limitations: MLP decoders are "lobotomized" due to poor scaling.
🔗 Resources:
- Original post ↗
- Original source ↗
- NNTC ↗
- Brief description of NNTC weight scaling limitations
🎤 Music - Chaos EP Idea
The author is struggling to find a decent rapper who loves Tzeentch, a chaos god, to collaborate on their EP. Two potential collaborators dropped out after waiting for six months each, leaving the demo over a year old.
Key Points:
Collaboration Challenges: The author is struggling to find a decent rapper who loves Tzeentch.
Collaborator Dropouts: Two potential collaborators dropped out after waiting for six months each.
Demo Age: The demo is over a year old.
🔗 Resources:
- Original post ↗
- Original source ↗
- Tzeentch ↗
- Brief description of the chaos EP idea
🤖 AI/ML - Jev and TabFM
If you're blown away by Jev from a philosophical perspective, the author recommends looking into Google's TabFM and its credited lineage, aka intelligence in spreadsheet.
Key Points:
Jev and TabFM: Jev is related to TabFM, which is a form of intelligence in spreadsheet.
TabFM Lineage: TabFM has a credited lineage in the field of intelligence in spreadsheet.
Philosophical Implications: Jev has philosophical implications that are worth exploring.
🔗 Resources:
- Original post ↗
- Original source ↗
- TabFM ↗
- Brief description of Jev and TabFM
🎨 Art - Video Analyzer
The author has released a video-analyzer tool that creates prompts for MiniMax H3 or LTX-2.5. The tool is available for use.
Key Points:
Video Analyzer: The author has released a video-analyzer tool.
Prompt Generation: The tool generates prompts for MiniMax H3 or LTX-2.5.
Tool Availability: The tool is available for use.
🔗 Resources:
- Original post ↗
- Original source ↗
- Video Analyzer ↗
- Brief description of the video-analyzer tool
🤖 AI/ML - Clairvoyance Guided Tour
The author has created a guided tour of Clairvoyance, a tool for creators. The tour is available for use.
Key Points:
Clairvoyance Tour: The author has created a guided tour of Clairvoyance.
Creator Tool: Clairvoyance is a tool for creators.
Tour Availability: The tour is available for use.
🔗 Resources:
- Original post ↗
- Original source ↗
- Clairvoyance ↗
- Brief description of the Clairvoyance guided tour
🚫 Politics - Bengaluru Taxation
The author is criticizing the Bengaluru government's taxation policies, which they claim are opaque and unfair. The author is advocating for transparency and accountability in taxation.
Key Points:
Taxation Criticism: The author is criticizing the Bengaluru government's taxation policies.
Transparency and Accountability: The author is advocating for transparency and accountability in taxation.
Taxation Opacity: The author claims that the taxation policies are opaque.
🔗 Resources:
- Original post ↗
- Original source ↗
- Bengaluru Taxation ↗
- Brief description of the taxation criticism
🤖 AI/ML - Cooperative Vectors
Cooperative Vectors is an upcoming extension that exposes neural-inference hardware to shaders. This does not require developers to use MLPs (neural nets with activations).
Key Points:
Cooperative Vectors: Cooperative Vectors is an upcoming extension.
Neural-Inference Hardware: The extension exposes neural-inference hardware to shaders.
MLP Independence: The extension does not require developers to use MLPs.
🔗 Resources:
- Original post ↗
- Original source ↗
- Cooperative Vectors ↗
- Brief description of Cooperative Vectors
🤖 AI/ML - Jev Performance
Jev performs well in some evals, but its local equivalent, Gemma, is worse and bloats the local app. The author is exploring shipping Jev locally.
Key Points:
Jev Performance: Jev performs well in some evals.
Gemma Comparison: Gemma is worse and bloats the local app.
Local Shipping: The author is exploring shipping Jev locally.
🔗 Resources:
- Original post ↗
- Original source ↗
- Jev ↗
- Brief description of Jev performance
🎨 Art - Moodboard Chat
The author has built a tiny Pinterest for AI, called Moodboard Chat, with @AdamSvystun. The tool allows users to explore moodboards, remix images, and feed their taste to their LLM.
Key Points:
Moodboard Chat: The author has built a tiny Pinterest for AI.
Collaboration: The tool was built with @AdamSvystun.
Tool Features: The tool allows users to explore moodboards, remix images, and feed their taste to their LLM.
🔗 Resources:
- Original post ↗
- Original source ↗
- Moodboard Chat ↗
- Brief description of Moodboard Chat
