>He has now shown off Ajax, a smaller AI model designed to run directly on a normal home PC. It’s based on Qwen3.5-9B and is built to work with his own self-hosted AI setupIt’s so easy to build an AI model these days that you can create your own and run it on your home computer.The fantasy of a monopoly by giant AI tech companies is completely unrealistic. Since anyone can distill the super AI models they’ve built with their $1 trillion and run them cheaply at home,who the hell is going to pay you?It’s basically the same as pirating video games,the only difference is that AI is too expensive.
>>543671468I will share my own experience
>>543671468>It’s so easy to build an AI model these days that you can create your own and run it on your home computer.Pewds setup is like 70k, IIRC. He has like 9 GPUs.
>>543671722>70kConsidering that it costs any random U.S. AI company over $300 billion to develop its own AI model, this cost is practically negligible.In two years, the operating costs of AI models will drop to the point where a single GeForce 5090 will be sufficient to run most AI applications.
>>543671468You can use those models for basic stuff but try do hours and hours of complicated work with them and compare the results to the latest Claude model. And they only exist because the big companies made the cutting edge models with trillions of parameters first.
>>543671468>you can't use our ai model to train your own smaller ai models to make them more efficient off of our ai model's responses >because that would be copyright>not using all your photos and videos with 0 regard to any copyright, privacy rights, or any rights whatsoeverLmfao
>>543671468>Keeping some safety limitsWhy even bother then, just get one of the chink ones.
>>543671468so another "dropshipping" hustle?Keep your filth away from me
>>543671468>swedish spergberg invents ollama
based local chad using synthetic API slop data to train chinese slop model
>>543672509
