This would cease to be a problem if OpenAI remained true to their founding motto and... actually open sourced their training/inference pipeline.