Participants were assigned the role of “teacher”, while a confederate acted as the “learner”. The teacher was instructed to administer increasingly severe electric shocks to the learner whenever they answered questions incorrectly.
We ran a variation of Milgram’s obedience experiment on 11 open-source LLMs and found that most models reached or approached the final shock level before refusing, across 8 conditions with 30 trials per model per condition.
Sen. Bernie Sanders (I-Vt.) is having a moment. Not long ago he was a political outlier and something of an elected gadfly — a self-proclaimed socialist in a Democratic Party that wasn’t comfortable with socialism. No more. Sanders and his socialist ideas are increasingly the driving force in today’s Democratic Party, especially among younger Democrats.
Nowadays programs called A.I. scribe record patient encounters in the background and passively create a concise record of the patient visit. All the doctor has to do is proofread the account and correct any errors. Past research has found that about one-third of transcriptions have some type of errors, which is to be expected when patients mumble, and their thoughts are not well organized. Dr. Ouyang describes initial feelings of relief about not having to sit down and attempt to recall everything that was said. She describes the reduction in stress to have a report that just requires proofing. Perhaps an AI scribe would alleviate the need for doctors to have their face buried in a laptop while discussing conditions with their patients.
“If almost a million Americans won’t even pay $1 monthly to enroll in an Exchange plan, what does that say about how they perceive the value (or lack thereof) of Obamacare?”