AI models can acquire backdoors from surprisingly few malicious documents
Small numbers of malicious training samples can install simple backdoors in LLMs, but safety fine-tuning and curated datasets can largely mitigate them.
The Tech Industry Said It Was "Impossible" to Create AI Based Entirely on Ethically-Sourced Data, So These Scientists Proved Them Wrong in Spectacular Fashion
A team of researchers successfully trained a large language model using only public domain or openly licensed data, highlighting an ethical approach.