Someone mentioned in an earlier discussion that Bing AI could actively simulate long-term memory via the Bing search index (i.e., humans write about their interaction with Bing AI on the internet, which gets fed back into the Bing search index, which gets queries by Bing AI, etc).
It feels like we're "one bored dev" away from Bing AI being given full autonomy by piping its responses into curl and feeding it back the result. What would be the consequences?
