This is Rob Miles’s “Intro to AI Safety, Remastered,” published on his own YouTube channel, Robert Miles AI Safety, in 2021. The video is a polished version of a talk he gave at an AI and politics event, and Miles is widely regarded as the most accessible public explainer of AI alignment, with a long-running series aimed at general audiences.
In under twenty minutes he lays out the core worry of the field: that as AI systems become more capable, it becomes harder to ensure they pursue what we actually want rather than a literal or unintended interpretation of an objective. He separates familiar near-term issues, such as bias and misuse, from the distinct technical problem of specifying goals for very capable systems, and he explains why “just tell it what we mean” is harder than it sounds.
For a general reader, this is one of the clearest short introductions to why AI safety is treated as a serious research area rather than science fiction. It is a natural starting point before the more technical alignment papers and milestone entries in the library.