I found this old video from Tom Scott and it seems like a rather good short dystopian introduction into the potential dangers of misalignment. Unlike other beginner resources, it makes no attempts as to explain technical terms or to give a comprehensive overview. Instead, Tom Scott simply describes a dystopian world and how a simple machine with a misaligned reward function can cause havoc from nothing very quickly.
Quite a few themes are touched upon namely extraordinarily vast growth in intelligence in short time spans, the perils of dealing with super-intelligence, the need for AI safety roles and oversight and the difficulty in setting goals.
Give it a watch and perhaps past it along the next time someone else is curious!
Bill
This is good, thanks! I’m not sure if it’s my preferred choice for any use case, but it’s close.
Thanks for sharing, will watch!