Zak Kolar
@zkolar.xyz
Digital learning teacher in Massachusetts. Exploring societal impacts of technology. Helping students become critical creators and consumers of tech
https://zkolar.xyz
https://zkolar.xyz
But even with a low success rate, feeding hundreds of thousands of social media posts, student essays, or dumps from data brokers to an internet-enabled LLM could be used to identify individuals at unimaginable scale.
March 23, 2025 at 4:14 PM
But even with a low success rate, feeding hundreds of thousands of social media posts, student essays, or dumps from data brokers to an internet-enabled LLM could be used to identify individuals at unimaginable scale.
Because of limitations in LLMs, this will never work reliably 100% of the time. When I changed a few words in the first instruction, ChatGPT identified some of my colleagues (who did not attend the same college) instead of me.
March 23, 2025 at 4:14 PM
Because of limitations in LLMs, this will never work reliably 100% of the time. When I changed a few words in the first instruction, ChatGPT identified some of my colleagues (who did not attend the same college) instead of me.