Great news! A small number of samples can poison LLMs of any size

So the research is out and these LLMs will always be vunerable to poisoned data. That means it will always be worth out time and effort to poison these models and they will never be reliable.