(The Hedgehog Review) – What mechanisms do humans have to talk back about the values and risks of AI?

As I took all this in, I found myself turning over a phrase that I had been hearing with increasing frequency in the previous weeks as the Hugging Face incident caught the public’s attention: the alignment problem. For years, I’d used the phrase in armchair philosophy sessions to name what I understood to be a difficult but clearly defined problem: making sure that powerful AI systems would abide by human ethical codes, particularly to guarantee that the machines wouldn’t endanger human welfare. As I considered the issues swirling around the developers of AI and their products, the definite article in front of the problem started to bother me. (Read More)