ChatGPT has gotten lots of hype on the internet in the last few days, which seems like a good opportunity to get a wider audience interested in AI Safety. It would be useful to have some nice short pitches / talking points for this, besides “Look how powerful AI is! Kind of scary”. I’d be interested to hear any suggestions / ideas that have worked well for people so far.
This seems to have gotten some attention on HN, albeit it's a more technical audience. https://news.ycombinator.com/item?id=33872062
I believe we should be developing AI Archetypes, similar to how Carl Jung aggregated his understanding of Analytical Psychology derived in how humans approach the world in repeating patterns of meta-stories.
I will write more about this as this is related to how can the alignment problem be approached from an evolutionary / psychological viewpoint..