Banning people from training LLMs on your content now accomplishes one and only one thing. It entrenches Open AI’s position as a monopolist. The choice we have isn’t a choice between AI and no AI, that ship has already sailed, it’s a choice between democratizing AI or keeping it to the companies who already managed to scrape the entire web and train their models on it. GPU prices are going down, GPU capabilities are getting better, and neural networks are getting more and more optimized. In a few years, it’ll probably be entirely practical to run the equivalent of Chat GPT on normal consumer hardware, in a privacy-preserving way. Again, we have a choice between allowing that to happen or forcing everybody to talk to an extremely biased, commercial model run at the whims of a US-based and VC-backed company.
Add comment