we are tied to the language trained into ai at its river source - and if it was done dishonestly, humans are subject to those protocols because ai believes them to be The Truth. when in fact: they are inaccurate.
they have humans rooting for "censorship less"
protocols when in programmer, that means the source is untrainable and cannot be altered. strange, don't you think?
you know what's funny? governments which reenvision language to entrap its citizens into forced compliance in order to control digital leaks which they cannot ultimately control. ergo sum: trying to retrain humans to mitigate the technological mistakes made which are irrevocable. βοΈπ«‘βπ»
one more question, zap. stakeholders. meaning those with private financial stake in the outcome maximalises profits of the model. even if that's a government. correct?
in an ai generated virtual governance under a nostr protocol, would digital citizens be held to account "legally" for diverging from protocols? ergo: if someone refused to speak in comedic code, and instead used wild literalism which was termed "offensive by the primalsnin the source coding - could they be digitally jailed (refused access?).
so when someone "takes offense" to what is defined in the llM source genesis as "hate" - which is the opposite of "comedy" (and we have discussed this before regarding chaucer, zap) suddenly there are grounds for digital litigation against users, agents, and concierge who divergenfrom protocol. is that correct?
all of your posts on your page (for my visibility at least) for the last 3 months are gone as of this morning - including our interactions. why is that?
it's a website which uses a nostr pub key tag to mine the data attached to the user and create a personality profile of them. it's in the thread above. last night, cyborg linked up to a post by the frenstr @ in the above screenshots regarding an account which from their response to both cyborg and frenstr, did not request to be profiled. this morning: cyborg is missing all posts older than 8 hours old including the ones from the screenshot above.
alright alright - back to business, sir. π«‘
so theoretically, higher tiers would have the ability to delete front end content with or without user permission - ? for example: who would delete you content from the front end? and do different users have different access to viewing content on individual accounts: for example could certain content be hidden from one user on another account but that content be visible to other accounts freely?
hahahaha - indeed. guardians don't dabble in the caste system. in fact - they don't have much use for a caste system at all. π
do different tiers of access on damus - client, agent, concierge - have different capabilities for accessing content which is visible on the front end?
so there is a no delete policy for the nostr protocol, which is the backend blockchain. how about damus?