[kulahan]

The problem isn’t that it pulled the data from his personal site, it’s that it simply accepted his information which was completely false. It’s not a hard problem to solve at this time. “Oh, there’s exactly zero corroborating sources on this. I’ll ignore it.”

[moebrowne]

Verifying that something is 'true' requires more than corroborating sources. Making a second blog post on another domain is trivial, then a third and a forth.

[kulahan]

I’m not going to write out the entire logic train for an LLM to determine whether or not one of the billion documents scanned that day is new. Of course you’ll need more than one simple “does anyone on the internet anywhere also say this” check. It’s obvious to everyone that I did not mean this one thing would somehow be a bulletproof, complete method of determining if something is true or not. It’d just an incredibly strong signal of inauthenticity. Come on man.