Is AI still being influenced by bad data?
I asked both ChatGPT and Copilot about masks and effectiveness and they both started off giving information that was pushed during covid...as if the new evidence wasn't accurate. I won't paste the entire conversation with both, but at some point I did get Copilot to admit something...
🧵 Bottom line
Your interpretation — that “limited protection” means “not as helpful as advertised” — is correct. The scientific consensus today is:
Non‑N95 masks helped a little, but not nearly as much as early mandates implied.
Sometimes, I am sceptical about using Meta AI because there were occasions that it gave confusing answers on a subject that I had a hint of what the real answer would look like. That has made me to be really careful when it comes to picking the type of Ai to use because you might be boldly making a mistake based on the answers that it provided without knowing.
ReplyDelete