Comments by "LoneTech" (@0LoneTech) on "Lawyer in Trouble for Filing ChatGPT-Created Fake Citations in Court" video.

  1. Note: Asking ChatGPT to check its work is not making ChatGPT check its work. ChatGPT is far lazier than you, and will blandly assume its own output was correct for no more reason than having said it.
    1
  2. No. Much more terrified that people are selling a product well known to hallucinate (it does not understand falsehood or knowing, so cannot knowingly tell falsehood) as panacea and more people don't even listen to the caveats, just preach and use it as infallible. The concern is reasonable but this system is far more basic.
    1
  3. Nope, that's not it. It's not conscious, and does not know or care what is true or what truth is. What you should fear is how many humans blindly believe it, or other bullshitters, over any evidence. For instance, be concerned about why ChatGPT is trained to advertise and advocate ChatGPT (note: not itself, it's not self aware). Its bias is way out of proportion with its performance, and that's because people put it there. However, you are correct that deception will be a greater problem once machine learning does begin to contain intelligence. It's already a common issue.
    1