👀9⚡2❤🔥1🏆1
Very incomplete,
But interesting breakdown idea.
Yes,
Definitely most of those most “in the culture” have very few followers, little engagement — Got to dig deep into the culture to find those next $BOMEs
Definitely many of the supposed tech influencers are horribly out of the culture, or opposite of the latest culture, in every way, e.g. Vitalik
And yeah, lots of the high-engagement ones that normies love are horribly disconnected both from the culture and the tech
Missing many influencers,
But nailed the principles here
🄳🄾🄾🄼🄿🄾🅂🅃🄸🄽🄶
But interesting breakdown idea.
Yes,
Definitely most of those most “in the culture” have very few followers, little engagement — Got to dig deep into the culture to find those next $BOMEs
Definitely many of the supposed tech influencers are horribly out of the culture, or opposite of the latest culture, in every way, e.g. Vitalik
And yeah, lots of the high-engagement ones that normies love are horribly disconnected both from the culture and the tech
Missing many influencers,
But nailed the principles here
🄳🄾🄾🄼🄿🄾🅂🅃🄸🄽🄶
👌5💯4
Smart thing about these new AI goonbots is that they keep their replies extremely short
— Which is enough to evade essentially all of the off-the-shelf AI detectors available today
But that’s a big problem, since platforms like Twitter and Telegram, where replies tend to be way too short for those AI detectors to judge
So this is why our new scoreboard AI system took an entirely different approach
— Which is essentially calculating the amount of information added in the reply vs what it’s replying to, using some AI techniques
= These would all be not counted, despite being way too short for classic AI detectors, because they add essentially zero information vs what they were replying to
So that’s why we did our anti-AI-spambot detection scoreboards in a whole new way.
Preliminary results — seems to work shockingly well at distinguishing AI replies from human replies, especially when looking at many replies made by an account.
🄳🄾🄾🄼🄿🄾🅂🅃🄸🄽🄶
— Which is enough to evade essentially all of the off-the-shelf AI detectors available today
But that’s a big problem, since platforms like Twitter and Telegram, where replies tend to be way too short for those AI detectors to judge
So this is why our new scoreboard AI system took an entirely different approach
— Which is essentially calculating the amount of information added in the reply vs what it’s replying to, using some AI techniques
= These would all be not counted, despite being way too short for classic AI detectors, because they add essentially zero information vs what they were replying to
So that’s why we did our anti-AI-spambot detection scoreboards in a whole new way.
Preliminary results — seems to work shockingly well at distinguishing AI replies from human replies, especially when looking at many replies made by an account.
🄳🄾🄾🄼🄿🄾🅂🅃🄸🄽🄶
👏6💯3❤🔥2