from Hacker News

OpenAI's o3 model is smart, but has a serious problem with lying

by eigenvalue on 4/25/25, 4:12 PM with 2 comments

• by jqpabc123 on 4/25/25, 4:21 PM

  Same stuff, different day --- real "truth" or "understanding" simply can't be obtained from a statistical model in any sort of reliable manner.

  As my professors used to say back in college, this should be intuitively obvious.