Genesis Startup Competition Pitch: NeuroMesh
summary
Claims a distributed training network that stitches many consumer GPUs into larger models using predictive coding networks as the core algorithm, showing proof-of-concept on small models with comparable loss to backprop and training speed approaching parity. Builds a web3-style marketplace using a token-incentivized GPU supply and demand model, targets cheaper heterogenous GPUs (RTX 30/40 series) rather than H100/A100, and plans a QC product launch with partners supplying secondhand and miner GPUs.
AI-written summary · about
related talks
13mDemocratizing AI with Open Source Compute
20mDecentralizing Compute for Machine Learning
52mSimplifying the adoption of generative AI for enterprises
23mFrom Gaming to the Metaverse
53mHow to Build and Scale Ultra-Lean Gen AI Startups
12mHow To Run AI At Scale Without Going Broke
57mBuilding GPU-accelerated multi-agent apps
05mFrom Open Source to Real-World Scale