4 ms·Thoughts on AI safety in this era of increasingly powerful open source LLMs3 points by duck 4y ago